A Brief Review of Elasticity and Viscoelasticity For Solids: Eview Rticle

Advances in Applied Mathematics and Mechanics DOI: 10.4208/aamm.
10-m1030
Adv. Appl. Math. Mech., Vol. 3, No. 1, pp. 1-51 February 2011
REVIEW ARTICLE
A Brief Review of Elasticity and Viscoelasticity
for Solids
Harvey Thomas Banks 1, , Shuhua Hu 1 and Zackary R. Kenz 1
1Center for Research in Scientific Computation and Department of Mathematics,
North Carolina State University, Raleigh, NC 27695-8212, USA
Received 25 May 2010; Accepted (in revised version) 31 August 2010
Available online 15 October 2010
Abstract. There are a number of interesting applications where modeling elastic

and/or viscoelastic materials is fundamental, including uses in civil engineering,
the food industry, land mine detection and ultrasonic imaging. Here we provide
an overview of the subject for both elastic and viscoelastic materials in order to
understand the behavior of these materials. We begin with a brief introduction of
some basic terminology and relationships in continuum mechanics, and a review
of equations of motion in a continuum in both Lagrangian and Eulerian forms. To
complete the set of equations, we then proceed to present and discuss a number
of specific forms for the constitutive relationships between stress and strain pro-
posed in the literature for both elastic and viscoelastic materials. In addition, we
discuss some applications for these constitutive equations. Finally, we give a com-
putational example describing the motion of soil experiencing dynamic loading by
incorporating a specific form of constitutive equation into the equation of motion.
AMS subject classifications: 93A30, 74B05, 74B20, 74D05, 74D10
Key words: Mathematical modeling, Eulerian and Lagrangian formulations in continuum me-
chanics, elasticity, viscoelasticity, computational simulations in soil, constitutive relationships.
1 Introduction
Knowledge of the field of continuum mechanics is crucial when attempting to under-
stand and describe the behavior of materials that completely fill the occupied space
and thus act like a continuous medium. There are a number of interesting applications
where modeling of elastic and viscoelastic materials is fundamental. One interest in
particular is in describing the response of soil which experiences some sort of impact.
Corresponding author.
URL: http://www.ncsu.edu/crsc/htbanks/
Email: htbanks@ncsu.edu (H. T. Banks), shu3@ncsu.edu (S. H. Hu), zrkenz@ncsu.edu (Z. R. Kenz)
http://www.global-sci.org/aamm 1 2011
c Global Science Press
2 H. T. Banks, S. H. Hu and Z. R. Kenz / Adv. Appl. Math. Mech., 3 (2011), pp. 1-51
This may result from buildings falling or being imploded, or even an intentionally in-
troduced impact as part of land mine detection efforts (see [54, 56]). The chief interest
is in determining what would happen to buried objects given a particular surface im-
pact. In the case of a building implosion, there are concerns for buried infrastructure
such as tunnels, pipes, or nearby building infrastructure. Investigations can be carried
out to determine the likely forces on these buried objects to ensure that the force de-
livered into the soil will not damage other infrastructure. When detecting land mines,
the methodology developed in the papers cited uses an impact on the ground to cre-
ate Rayleigh surface waves that are subsequently changed upon interacting with a
buried mine; this change in wave form might be detected through electromagnetic or
acoustic means. Creating a model that accurately describes these Rayleigh waves is
key to modeling and understanding the buried land mine situation. In both of these
examples, one must study the soil properties, determine a valid constitutive relation-
ship for the soil, and verify the accuracy of the model. One can then use the models
to predict the results from different forces, soil properties, etc. Another application is
the non-invasive detection of arterial stenosis (e.g., see [1, 3, 11, 38, 50]). In this study,
blockages in the artery create turbulence in the blood flow, which then generates an
acoustic wave with a normal and shear component. The acoustic wave propagates
through the chest cavity until it reaches the chest wall, where a series of sensors detect
the acceleration of the components of the wave. The data from the sensors can then be
used to quickly determine the existence and perhaps the location of the blockages in
the artery. This technique is inexpensive and non-invasive. For such a technology to
be feasible, a mathematical model that describes the propagation of the acoustic wave
from the stenosis to the chest wall will be necessary to correctly detect the location of
a blockage.
The goal of this paper is to provide a brief introduction of both elastic and vis-
coelastic materials for those researchers with little or no previous knowledge on con-
tinuum mechanics but who are interested in studying the mechanics of materials. The
materials that we are considering are simple (for example the stress at a given material
point depends only on the history of the first order spatial gradient of the deforma-
tion in a small neighborhood of the material point and not on higher order spatial
gradients) and non-aging (the microscopic changes during an experiment can be ne-
glected in the basic model). Our presentation is part tutorial, part review but not a
comprehensive survey of a truly enormous research literature. We rely on parts of the
standard literature and discuss our view of generally accepted concepts. We present
a discussion of topics we have found useful over the past several decades; hence ap-
proximately 20% of references are work from our group. We have not meant to ignore
major applications in the many fine contributions of others; rather our presentation
reflects a certain level of comfort in writing about efforts on which we have detailed
knowledge and experience.
The introductory review is outlined as follows: in Section 2 some basic terminol-
ogy (such as stress and strain) and relationships (e.g., the relationship between strain
and displacement) of continuum mechanics are briefly described. In addition, we give
H. T. Banks, S. H. Hu and Z. R. Kenz / Adv. Appl. Math. Mech., 3 (2011), pp. 1-51 3
a brief overview of equations that describe the general motion of a continuous material
in both Eulerian and Lagrangian forms which describe motion utilizing a general rela-
tionship between stress and strain for materials. We thus must introduce constitutive
relationships (see Section 3) in order to quantify the material dependent relationship
between stress and strain. A wide number of approaches to model these constitutive
relationships have been developed and we focus much of our attention here on these.
Once constitutive relationships are determined one can in principle solve the resulting
set of motion equations with proper boundary and initial conditions. We conclude the
paper in Section 4 by giving an example describing the motion of soil experiencing
dynamic loading.
We remark that all of the considerations here are under isothermal conditions; in
most physical problems where energy considerations (heat flow, temperature effects,
entropy, etc.) are important, one may need to treat thermoelastic/thermoviscoelastic
modeling. An introduction to this more challenging theory can be found in Chapter
III of [16] and Chapter 5 of [30] with a more sophisticated thermodynamic treatment
in [60]. We do not give details here since the subject is beyond the scope of our review.
2 Preliminary notions and balance laws

Throughout this review, bold letters are used to denote vectors unless otherwise indi-
cated, | | is used to denote the determinant of a matrix, and ij denotes the Kronecker
delta, that is, ij = 1 for i = j and zero otherwise. For convenience of presentation, we
may occasionally use the Einstein notational convention (or index convention), where
the convention is as follows: the repetition of an index in a term will denote a summa-
tion with respect to that index over its range. For example, the Einstein representations
for k l Cijkl kl and i j ij dxi dx j are given by Cijkl kl and ij dxi dx j , respectively. In
the Cartesian coordinate system we may denote coordinate axes by x, y, and z or by x1 ,
x2 and x3 , depending again on the ease of presentation. Accordingly, the components
of a tensor are denoted by xx , xy , xz , etc., in reference to coordinates x, y and z,
and are denoted by ij , i, j = 1, 2, 3 in reference to coordinates x1 , x2 and x3 .
In this section we first introduce some basic terminology and relationships used
in continuum mechanics and then give a review on some fundamental physical laws
such as the conservation of mass, and equation of motions in both Lagrangian and
Eulerian forms. The content of this section is a summary of material from several
popular mechanics books including [24, 26, 39, 44].
2.1 Preliminaries
2.1.1 Kinematics: deformation and motion
An external force applied to an object results in a displacement, and the displacement
of a body generally has two components:
1. A rigid-body displacement: in this case the relative displacement between par-

ticles is zero, i.e., the shape and size of the body does not change.
2. A deformation: here there is a relative displacement between particles, i.e, the
shape and/or the size are changed. There are two formulations describing deforma-
tion:
i. Finite strain theory which deals with deformations in which both rotations and strains
are arbitrarily large. For example, elastomers, plastically-deforming materials and other fluids
and biological soft tissue often undergo large deformations and may also require viscoelastic
ideas and formulations.
ii. Infinitesimal strain theory which treats infinitesimal deformations of a continuum
body. Many materials found in mechanical and civil engineering applications, such as con-
crete and steel, typically undergo small deformations.
For this presentation, we shall focus on deformations. When analyzing the de-
formation or motion of solids, or the flow of fluids, it is traditional (and helpful) to
describe the sequence or evolution of configurations throughout time. One descrip-
tion for motion is made in terms of the material or fixed referential coordinates, and
is called a material description or the Lagrangian description. The other description for
motion is made in terms of the spatial or current coordinates, called a spatial descrip-
tion or Eulerian description. An intuitive comparison of these two descriptions would
be that in the Eulerian description one places the coordinate or reference system for
motion of an object on the object as it moves through a moving fluid (e.g., on a boat in
a river) while in the Lagrangian description one observes and describes the motion of
the object from a fixed vantage point (e.g., motion of the boat from a fixed point on a
bridge over the river or on the side of the river.).
Lagrangian description
In a Lagrangian description an observer standing in the referential frame observes the
changes in the position and physical properties as the material particles move in space
as time progresses. In other words, this formulation focuses on individual particles as
they move through space and time. This description is normally used in solid mechanics.
In the Lagrangian description, the motion of a continuum is expressed by the mapping
function h given by
x = h(X, t), (2.1)
which is a mapping from initial (undeformed/material) configuration 0 to the present
(deformed/spatial) configuration t . Hence, in a Lagrangian coordinate system the
velocity of a particle at X at time t is given by
x h(X, t)
V(X, t) = = ,
t t
and the total derivative (or material derivative) of a function (X, t), which is denoted
by a dot or the symbol D/Dt, is just the partial derivative of with respect to t,
D
(X, t) = (X, t).
Dt t
Eulerian description
An Eulerian description focuses on the current configuration t , giving attention to
what is occurring at a moving material point in space as time progresses. The coordi-
nate system is relative to a moving point in the body and hence is a moving coordinate
system. This approach is often applied in the study of fluid mechanics. Mathematically, the
motion of a continuum using the Eulerian description is expressed by the mapping
function
X = h1 (x, t),
which provides a tracing of the particle which now occupies the position x in the
current configuration t from its original position X in the initial configuration 0 .
The velocity of a particle at x at time t in the Eulerian coordinate system is
( )
v(x, t) = V h1 (x, t), t .
Hence, in an Eulerian coordinate system the total derivative (or material derivative)
of a function (x, t) is given by
D 3

(x, t) = (x, t) + vi (x, t) = (x, t) + v(x, t) (x, t).
Dt t i =1
xi t
Remark 2.1. There are a number of the different names often used in the literature to
refer to Lagrangian and Eulerian configurations. Synonymous terminology includes
initial/referential, material, undeformed, fixed coordinates for Lagrangian and cur-
rent/present, space, deformed, moving coordinates for Eulerian reference frames.
2.1.2 Displacement and strain

A particle P located originally at the coordinate X = ( X1 , X2 , X3 ) T is moved to a place
P with coordinate x = ( x1 , x2 , x3 )T when the body moves and deforms. Then the vec-
tor PP , is called the displacement or deformation vector of the particle. The displacement
vector is
x X. (2.2)
Let the variable X = ( X1 , X2 , X3 ) T identify a particle in the original configuration
of the body, and x = ( x1 , x2 , x3 ) T be the coordinates of that particle when the body
is deformed. Then the deformation of a body is known if x1 , x2 and x3 are known
functions of X1 , X2 , X3 :
x i = x i ( X 1 , X2 , X3 ) , i = 1, 2, 3.
The (Lagrangian) displacement of the particle relative to X is given by
U(X) = x(X) X. (2.3)
If we assume the transformation has a unique inverse, then we have
Xi = Xi ( x 1 , x 2 , x 3 ) , i = 1, 2, 3,
Figure 1: Deformation of a body.
for every particle in the body. Thus the (Eulerian) displacement of the particle relative
to x is given by
u ( x ) = x X ( x ). (2.4)
To relate deformation with stress, we must consider the stretching and distortion of
the body. For this purpose, it is sufficient if we know the change of distance between
any arbitrary pair of points.
Consider an infinitesimal line segment connecting the point P( X1 , X2 , X3 ) to a
neighboring point Q( X1 + dX1 , X2 + dX2 , X3 + dX3 ) (see Fig. 1). The square of the
length of PQ in the original configuration is given by
|dX|2 = (dX)T dX = (dX1 )2 + (dX2 )2 + (dX3 )2 .
When P and Q are deformed to the points P ( x1 , x2 , x3 ) and Q ( x1 + dx1 , x2 + dx2 , x3 +
dx3 ), respectively, the square of length of P Q is
|dx|2 = (dx)T dx = (dx1 )2 + (dx2 )2 + (dx3 )2 .
Definition 2.2. The configuration gradient (often, in something of a misnomer, referred to as
deformation gradient in the literature) is defined by

x1 x1 x1
X1 X2 X3
dx
A= =

x2
X1
x2
X2
x2
X3
.
(2.5)
dX x3 x3 x3
X1 X2 X3
The Lagrangian strain tensor

The Lagrangian strain tensor is measured with respect to the initial configuration (i.e.,
Lagrangian description). By the definition of configuration gradient, we have dx =
AdX and
|dx|2 |dX|2 =(dx)T dx (dX)T dX
=(dX)T A T AdX (dX)T dX
=(dX)T ( A T A I )dX.
The Lagrangian (finite) strain tensor E is defined by

1 T
E= ( A A I ). (2.6)
2
The strain tensor E was introduced by Green and St. Venant. Accordingly, in the liter-
ature E is often called the Greens strain tensor or the Green-St. Venant strain tensor. In
addition, from (2.6) we see that the Lagrangian strain tensor E is symmetric.
If the strain satisfies A T A I = 0, then we say the object is undeformed; otherwise,
it is deformed. We next explore the relationship between the displacement and strain.
By (2.3) and (2.5) we have the true deformation gradient given by
x1 x1 x1
X1 1 X2 X3

U = X1
x2 x2
X2 1 x2
X3
= A I,

x3 x3 x3
X1 X2 X3 1
or
Ui xi
= ij , j = 1, 2, 3, i = 1, 2, 3,
X j X j
where I is the identity matrix, and U = (U1 , U2 , U3 ) T . Thus, because
A = U + I,
the relationship between Lagrangian strain (2.6) and displacement is given by

1 [ Ui Uj Uk Uk ]
Eij = + + ,
2 X j Xi Xi X j
where Eij is the (i, j) component of strain tensor E.
The Eulerian strain tensor

The Eulerian strain tensor is measured with respect to the deformed or current config-
uration (i.e., Eulerian description). By using dX = A1 dx, we find
|dx|2 |dX|2 =(dx)T dx (dx)T ( A1 )T A1 dx

=(dx)T ( I ( A1 )T A1 )dx,
and the Eulerian (finite) strain tensor e is defined by
1( )
e= I ( A 1 ) T A 1 . (2.7)
2
The strain tensor e was introduced by Cauchy for infinitesimal strains and by Almansi
and Hamel for finite strains; e is also known as Almansis strain in the literature. In
addition, we observe from (2.7) that Eulerian strain tensor e is also symmetric.
We also may give the relationship between the displacement and strain in the Eu-
lerian formulation. By (2.4) and (2.5) we have
X1
1 x1 X
x2
1
X
x3
1

u =

X2
x1 1 X2
x2 X
x3
2 = I A 1 ,

X
x1
3 X3
x2 1 X3
x3
or
ui Xi
= ij , j = 1, 2, 3, i = 1, 2, 3,
x j x j
where u = (u1 , u2 , u3 )T . Thus, the relationship between Eulerian strain and displace-
ment is given by
1 [ ui u j u u ]
eij = + k k ,
2 x j xi xi x j
where eij is the (i, j) component of strain tensor e.
Remark 2.3. There are two tensors that are often encountered in the finite strain theory.
One is the right Cauchy-Green configuration (deformation) tensor, which is defined by
( x x )
k k
DR = A T A = ,
Xi X j
and the other is the left Cauchy-Green configuration (deformation) tensor defined by
( x x )
i j
DL = AA T = .
Xk Xk
The inverse of DL is called the Finger deformation tensor. Invariants of DR and DL are
often used in the expressions for strain energy density functions (to be discussed below
in Section 3.1.2). The most commonly used invariants are defined to be the coefficients
of their characteristic equations. For example, frequently encountered invariants of DR
are defined by
I1 = tr( DR ) = 21 + 22 + 23 ,
1[ ]
I2 = tr( DR2 ) (tr( DR ))2 = 21 22 + 22 23 + 23 21 ,
2
I3 = det( DR ) = 21 22 23 ,
where i , i = 1, 2, 3 are the eigenvalues of A, and also known as principal stretches

(these will be discussed later in Section 3.1.2).
Infinitesimal strain theory

Infinitesimal strain theory is also called small deformation theory, small displacement the-
ory or small displacement-gradient theory. In infinitesimal strain theory, it is assumed that
the components of displacement ui are such that their first derivatives are so small that
higher order terms such as the squares and the products of the partial derivatives of ui
are negligible compared with the first-order terms. In this case eij reduces to Cauchys
infinitesimal strain tensor
1 [ ui u j ]
ij = + , (2.8)
2 x j xi
(hence the infinitesimal strain tensor is also symmetric). Thus, ij = ji . We note that
in this case the distinction between the Lagrangian and Eulerian strain tensors disap-
pears (i.e., Eij eij ij ), since it is immaterial whether the derivatives of the dis-
placement are calculated at the position of a point before or after deformation. Hence,
the necessity of specifying whether the strains are measured with respect to the initial
configuration (Lagrangian description) or with respect to the deformed configuration
(Eulerian description) is characteristic of a finite strain analysis and the two different
formulations are typically not encountered in the infinitesimal theory.
2.1.3 Stress
Stress is a measure of the average amount of force exerted per unit area (in units N/m2
or Pa), and it is a reaction to external forces on a surface of a body. Stress was intro-
duced into the theory of elasticity by Cauchy almost two hundred years ago.
Definition 2.4. The stress vector (traction) is defined by
dF F
T(n) = = lim ,
d 0
where the superscript (n) is introduced to denote the direction of the normal vector n of the
surface and F is the force on the surface.
To further elaborate on this concept, consider a small cube in the body as depicted
in Fig. 2 (left). Let the surface of the cube normal (perpendicular) to the axis z be
donated by z . Let the stress vector that acts on the surface z be T(e3 ) , where
e3 = (0, 0, 1)T . Resolve T(e3 ) into three components in the direction of the coordinate
axes and denote them by zx , zy and zz . Similarly we may consider surface x and
y perpendicular to x and y, the stress vectors acting on them, and their components
in the x, y and z directions. The components xx , yy and zz are called normal stresses,
and xy , xz , yx , yz , zx and zy are called shear stresses. A stress component is posi-
tive if it acts in the positive direction of the coordinate axes. We remark that the notation
face, direction is consistently used in elasticity theory.
Figure 2: Notations of stress components.
Definition 2.5. The Cauchy stress tensor is defined by

[ ] xx yx zx
= T(e1 ) T(e2 ) T(e3 ) = xy yy zy , (2.9)
xz yz zz
where
e1 = (1, 0, 0) T , e2 = (0, 1, 0)T , and e3 = (0, 0, 1) T .
We have the following basic formulation due to Cauchy.
Theorem 2.6. (see [25, pp.69]) Let T(n) be the stress vector acting on d whose outer normal
vector is n, as illustrated in Fig. 3. Cauchys formula expresses T(n) as a function of the stress
vectors on the planes perpendicular to the coordinate axes, i.e., in terms of the components of
the Cauchy stress tensor. This formula asserts that
(n)
Tx = xx n x + yx ny + zx nz , (2.10a)
(n)
Ty = xy n x + yy ny + zy nz , (2.10b)
(n)
Tz = xz n x + yz ny + zz nz . (2.10c)
Figure 3: Stress vector acting on a plane with normal n.

Here ( (n) (n) (n) ) T

T(n) = Tx , Ty , Tz , and n = (n x , ny , nz ) T .
Cauchys formula (2.10) can be written concisely as
T(n) = n,
where is the Cauchy stress tensor defined in (2.9).
Remark 2.7. In addition to the Cauchy stress tensor, there are other stress tensors
encountered in practice, such as the first Piola-Kirchhoff stress tensor and second Piola-
Kirchhoff stress tensor. The differences between the Cauchy stress tensor and the Piola-
Kirchhoff stress tensors as well as relationships between the tensors can be illustrated
as follows:
1. Cauchy stress tensor: relates forces in the present (deformed/spatial) configuration
to areas in the present configuration. Hence, sometimes the Cauchy stress is also called
the true stress. In addition, the Cauchy stress tensor is symmetric, which is implied
by the fact that the equilibrium of an element requires that the resultant moments
vanish. We will see in Section 2.2.4 that the Cauchy stress tensor is used in the Eulerian
equation of motion. Hence the Cauchy stress tensor is also referred to as the Eulerian
stress tensor.
2. First Piola-Kirchhoff stress tensor (also called the Lagrangian stress tensor in [24, 25]):
relates forces in the present configuration with areas in the initial configuration. The
relationship between the first Piola-Kirchhoff stress tensor P and the Cauchy stress
tensor is given by
P = | A | ( A 1 ) T . (2.11)
From the above equation, we see that in general the first Piola-Kirchhoff stress tensor
is not symmetric (its transpose is called the nominal stress tensor or engineering stress
tensor). Hence, the first Piola-Kirchhoff stress tensor will be inconvenient to use in a
stress-strain law in which the strain tensor is always symmetric. In addition, we will
see in Section 2.2.5 that the first Piola-Kirchhoff stress tensor is used in the Lagrangian
equation of motion. As pointed out in [25], the first Piola-Kirchhoff stress tensor is the
most convenient for the reduction of laboratory experimental data.
3. Second Piola-Kirchhoff stress tensor (referred to as Kirchoff stress tensor in [24, 25],
though in some references such as [44] Kirchhoff stress tensor refers to a weighted
Cauchy stress tensor and is defined by | A|): relates forces in the initial configura-
tion to areas in the initial configuration. The relationship between the second Piola-
Kirchhoff stress tensor S and the Cauchy stress tensor is given by
S = | A | A 1 ( A 1 ) T . (2.12)
From the above formula we see that the second Piola-Kirchhoff stress tensor is sym-
metric. Hence, the second Piola-Kirchhoff stress tensor is more suitable than the first
Piola-Kirchhoff stress tensor to use in a stress-strain law. In addition, by (2.11) and
(2.12) we find that
S = A1 P. (2.13)
Note that for infinitesimal deformations, the Cauchy stress tensor, the first Piola-
Kirchhoff stress tensor and the second Piola-Kirchhoff tensor are identical. Hence,
it is necessary only in finite strain theory to specify whether the stresses are measured
with respect to the initial configuration (Lagrangian description) or with respect to the
deformed configuration (Eulerian description).
2.2 Equations of motion of a continuum

There are two common approaches in the literature to derive the equation of motion of
a continuum: the differential equation approach (for example, in [26]) and the integral
approach (for example, in [26, 44]). In this section, we will use an integral approach to
derive the equation of continuity and the equation of motion of a continuum, first in
the Eulerian (or moving) coordinate system and then in the Lagrangian coordinate system.
In the following, we will sometimes use to denote t , and to denote t for ease in
the presentation; these will refer to volume or surface elements, respectively, that are
time dependent.
Before deriving the equations of motion of a continuum, we first discuss forces.
There are two types of external forces acting on material bodies in the mechanics of
continuum media:
1. Body forces (N/m3 ), acting on elements of volume of body. For example, gravi-
tational forces and electromagnetic forces are body forces.
2. Surface forces (N/m2 ), or stress, acting on surface elements. For example, aero-
dynamics pressure acting on a body, stress between one part of a body on another, etc.,
are surface forces.
Then the total force F acting upon the material occupying the region interior to
a closed surface is I
F = T(n) d + fd, (2.14)

where f = ( f x , f y , f z )T is the body force, and T(n) is the stress vector acting on d
whose outer normal vector is n.
The expression (2.14) is a universal force balance statement independent of any
particular coordinate system being used. Of course, with either the Eulerian or La-
grangian formulation, the stresses and forces must be expressed in terms of the appro-
priate coordinate system.
2.2.1 The material derivative of a volume integral

To carry out our derivations, we need a calculus for interchanging integration and
differentiation when both the limits of the integration and the integrand depend on
the differentiation variable. Let (t) be a volume integral of a continuously differen-
tial function ( x, y, z, t) defined over a spatial domain t occupied by a given set of
material particles at time t, i.e.,

(t) = ( x, y, z, t)dxdydz.
t
Then the rate of change of (t) with respect to t is given by (suppressing the multiple
integral notation here and below when it is clearly understood that the integral is a
volume or surface integral)

d
= d + (v x n x + vy ny + vz nz )d, (2.15)
dt t t t
where on the boundary t of t , v = v(t) is the velocity

( dx dy dz ) T
v(t) = , , .
dt dt dt
Eq. (2.15) can be written concisely as

d
= d + v nd.
dt t t t
The first term on the right side corresponds to rate of change in a fixed volume, and
the second term corresponds to the convective transfer through the surface. By Gauss
theorem, Eq. (2.15) can also be written as
(
d v x vy vz )
= + + + d, (2.16)
dt t t x y z
or more concisely as
( )
d
= + (v) d.
dt t t
This rate, called the material derivative of , is defined for a given set of material parti-
cles in a moving volume. We note that when t = 0 for all t (i.e., the boundary is
not moving so that v = 0), this becomes simply

d
( x, y, z, t)dxdydz = ( x, y, z, t)dxdydz.
dt 0 0 t
2.2.2 The equation of continuity
We next derive the equation of continuity for an arbitrary mass of particles that may
be changing in time. The mass contained in a domain t at time t is

m(t) = ( x, y, z, t)dxdydz.
t
Conservation of mass requires that dm/dt = 0 and thus we have from (2.16)
[
dm v x vy vz ]
= + + + d.
dt t t x y z
Hence, we obtain [ v x vy vz ]
+ + + d = 0.
t t x y z
Since the above equality holds for an arbitrary domain t , we obtain the pointwise
equation of continuity
v x vy vz
+ + + = 0, (2.17)
t x y z
which can be written concisely as

+ (v) = 0.
t
2.2.3 The Reynolds transport theorem

In this subsection, we will use the material derivative (2.16) as well as the equation of
continuity (2.17) to derive the celebrated Reynolds transport theorem. By (2.16) we find
that (
d (vz ) vz v x vz vy vz vz )
vz d = + + + d.
dt t t t x y z
Then using the equation of continuity (2.17), we find that the integrand of the right
side of the above equation is equal to
vz ( v vy vz ) vz vz vz
x
vz + + vz + + + v x + vy + vz
t t x y z x y z
( v vy vz ) ( vz vz vz vz )
x
=vz + + + + + vx + vy + vz
t x y z t x y z
( v vz vz vz )
z
= + vx + vy + vz .
t x y z
Hence, we have
( v
d z vz vz vz )
vz d = + vx + vy + vz d. (2.18)
dt t t t x y z
Eq. (2.18) is the Reynolds transport theorem, which is usually written concisely as

d Dvz
vz d = d,
dt t t Dt
where Dvz /Dt is the total derivative of vz , and is given by
D vz vz vz vz
vz ( x, y, z, t) = + vx + vy + vz .
Dt t x y z
We note that the above is independent of any coordinate system and depends only on
the rules of calculus and the assumptions of continuity of mass in a time dependent
volume of particles.
2.2.4 The Eulerian equations of motion of a continuum

We are now ready to use the above rules of calculus and the continuity of mass as
embodied in the Reynolds transport theorem to derive the equations of motion in
an Eulerian coordinate system. Throughout we have = t and = t (we will
suppress the subscripts) and we assume the coordinate system ( x, y, z) is now moving
(changing with the volume element) with a velocity v = (dx/dt, dy/dt, dz/dt)T of the
deformation of the material. The resultant force Fz in the z-direction on an arbitrary
volume is
(n)
Fz = Tz d + f z d. (2.19)

By Cauchys formula (2.10) and Gauss theorem we have
( )
(n)
Tz d = xz n x + yz ny + zz nz d

(
xz yz zz )
= + + d.
x y z
Hence, by the above equality and (2.19), we obtain
( )
xz yz zz
Fz = + + + f z d.
x y z
Newtons law states that
( )
d xz yz zz
vz d = + + + f z d.
dt x y z
Hence, by the Reynolds transport theorem we have
( v ( )
z vz vz vz ) xz yz zz
+ vx + vy + vz d = + + + f z d.
t x y z x y z
Note that because the above equality holds for an arbitrary domain , the integrands
on both sides must be equal. Thus, we have
( v vz vz vz ) xz yz zz
z
+ vx + vy + vz = + + + fz,
t x y z x y z
or written concisely as
Dvz
= ,z + f z ,
Dt
which is the equation of motion of a continuum in the z-direction. The entire set for the
equations of motion of a continuum in an Eulerian coordinate system is given as follows:
( v v x v x v x ) xx yx zx
x
+ vx + vy + vz = + + + fx, (2.20a)
t x y z x y z
( v vy vy vy ) xy yy zy
y
+ vx + vy + vz = + + + fy, (2.20b)
t x y z x y z
( v vz vz vz ) xz yz zz
z
+ vx + vy + vz = + + + fz. (2.20c)
t x y z x y z
We note that (2.20) is also called Cauchys equation of motion or Cauchys momentum
equation in some literature. Eq. (2.20) can be written in vector form as
( v )
+ (v )v = + f,
t
where is the Cauchy stress tensor defined in (2.9). It is often desirable to express
these equations of motion in terms of displacements u. We find (because the Eulerian
velocity is given in terms of the displacement (2.4) by v = u/t)
[ 2 u ( u ) u ]
+ = + f. (2.21)
t2 t t
2.2.5 The Lagrangian equations of motion of a continuum
Next we will rewrite (2.20) in terms of a Lagrangian description, that is, we will de-
rive an equation of motion in the Lagrangian coordinate system (O-XYZ coordinate
system). Let 0 denote the boundary of 0 in the initial (undeformed/material) con-
figuration, and n0 be the outer normal vector on 0 . By Nansons formula [44] we
have
nd = | A|( A1 ) T n0 d0 , (2.22)
where n0 = (n0X , n0Y , n0Z ) T , A is the configuration gradient defined by (2.5). Multi-
plying both sides of (2.22) by we obtain
nd = | A| ( A1 ) T n0 d0 .
By (2.11), we have
nd = Pn0 d0 .
Let f0 be the external body force acting on 0 (f0 = | A|f), let 0 ( X, Y, Z, t) be the ma-
terial density in the Lagrangian coordinate system (conservation of mass implies that
0 = | A|), and V ( X, Y, Z, t) be the velocity in the Lagrangian coordinate system.
Then we can rewrite the resultant force in the z-direction in the Eulerian coordinate
system as the resultant force in the Z direction in the Lagrangian coordinate system,
which is

F0Z = ( PZX n0X + PZY n0Y + PZZ n0Z )d0 + f 0Z d0 .
0 0
Then by Gauss Theorem and the above equation we find that

( P
ZX P P )
F0Z = + ZY + ZZ d0 + f 0Z d0 .
0 X Y Z 0
We can rewrite Reynolds transport theorem (2.18) in the Lagrangian coordinate system
and find
d Dvz DVZ
vz d = d = 0 d0 .
dt Dt 0 Dt
Note that
DVZ V
= Z.
Dt t
Hence, we can rewrite Newtons law in the Lagrangian coordinate system as
( P )
VZ ZX PZY P
0 0 = + + ZZ + f 0Z d0 .
0 t 0 X Y Z
Note that the above equality holds for any 0 . Thus we have
VZ P P P
0 = ZX + ZY + ZZ + f 0Z ,
t X Y Z
which is the equation of motion in the Z-direction. Then the equations of motion in the
Lagrangian coordinate system are given by
VX P P P
0 = XX + XY + XZ + f 0X , (2.23a)
t X Y Z
VY PYX PYY PYZ
0 = + + + f 0Y , (2.23b)
t X Y Z
VZ P P P
0 = ZX + ZY + ZZ + f 0Z , (2.23c)
t X Y Z
or, written concisely,
V
0 = P + f0 .
t
Note that
U
. V=
t
Hence, the Lagrangian equations of motion in terms of displacement is given by
2 U
0 = P + f0 . (2.24)
t2
Remark 2.8. We note that the equations of motion (2.21) in the Eulerian (or mov-
ing) coordinate system are inherently nonlinear independent of the constitutive law
assumptions (discussed in the next section) we might subsequently adopt. On the
other hand, the Lagrangian formulation (2.24) (relative to a fixed referential coordi-
nate system) will yield a linear system if a linear constitutive law is assumed. Thus,
there are obvious advantages to using the Lagrangian formulation in linear theory
(i.e., when a linear constitutive law is assumed).
3 Constitutive relationships: stress and strain

In the preceding discussions, we have focused on relationships between displace-
ments (and their rates) and the stress tensors. We have also related strain tensors
to displacements. To complete our derivations of the equations of motion, we must

know (or assume) the relationships (constitutive laws) between stress and strain.
Constitutive laws are usually formulated based on empirical observations, and they
hold for a given material and are thus material dependent. Moreover, they must
be independent of any referential coordinate system that we choose. In addition, we
must note that the constitutive law describes an ideal material, and it should provide
a close approximation to the actual behavior of the real material that this constitutive
law is intended to model.
The concept of isotropy is used frequently as a simplifying assumption in contin-
uum mechanics, and many useful materials are indeed isotropic or approximately so.
We proceed to present the formal definition of an isotropic tensor and isotropic mate-
rials.
Definition 3.1. If a tensor has the same array of components when the frame of reference is
rotated or reflected (i.e., invariance under rotation or reflection), then it is said to be an isotropic
tensor. A material whose constitutive equation is isotropic is said to be an isotropic material.
Remark 3.2. If the tensor Dijkl is isotropic, then it can be expressed in terms of two
independent constants and by
Dijkl = ij kl + (ik jl + il jk ). (3.1)
We note that here is a Lame parameter not to be confused with the Poisson ratio also
encountered in elasticity.
Since we are interested in incorporating the constitutive laws for stress and strain
into the equations of motion, we will only present constitutive laws for their relax-
ation forms, i.e., stress is a function of strain and/or strain rate. The corresponding
compliance forms, i.e., the strain in terms of stress and/or stress rate, for most of these
constitutive laws can be defined similarly by just interchanging the role of stress and
strain. For convenience, we will suppress the spatial dependence of both stress and
strain when the constitutive relationship is given. Recall also that in an infinitesimal
setting the stress tensors are all equivalent; unless noted otherwise, we will use to
denote the stress in the following discussion and assume an infinitesimal setting. The
rest of this section is outlined as follows: we first talk about the constitutive equations
used in elastic materials in Section 3.1, and then we present and discuss a number of
constitutive laws appearing in the literature for the viscoelastic materials in Section
3.2.
3.1 Elastic materials

Elasticity is the physical property of a material that when it deforms under stress (e.g.,
external forces), it returns to its original shape when the stress is removed. For an
elastic material, the stress-strain curve is the same for the loading and unloading pro-
cess, and the stress only depends on the current strain, not on its history. A familiar
example of an elastic material body is a typical metal spring. Below we will discuss
linear elasticity in Section 3.1.1 and then follow with comments on nonlinear elasticity
in Section 3.1.2.
3.1.1 Linear elasticity

The classical theory of elasticity deals with the mechanical properties of elastic solids
for which the stress is directly proportional to the stress in small deformations. Most
structural metals are nearly linear elastic under small strain and follow a constitutive
law based on Hookes law. Specifically, a Hookean elastic solid is a solid that obeys
Hookes Law
ij = cijkl kl , (3.2)
where cijkl is elasticity tensor. If a material is isotropic, i.e., the tensor cijkl is isotropic,
then by (3.1) and (3.2) we have
ij = ij kk + 2 ij , (3.3)
where and are called Lames parameters. In engineering literature, the second Lame
parameter is further identified as the shear modulus.
3.1.2 Nonlinear elasticity

There exist many cases in which the material remains elastic everywhere but the stress-
strain relationship is nonlinear. Examples are a beam under simultaneous lateral and
end loads, as well as large deflections of a thin plate or a thin shell. Here we will
concentrate on the hyperelastic (or Green elastic) material, which is an ideally elastic
material for which the strain energy density function (a measure of the energy stored
in the material as a result of deformation) exists. The behavior of unfilled, vulcanized
elastomers often conforms closely to the hyperelastic ideal.
Nonlinear stress-strain relations

Let W denote the strain energy function, which is a scalar function of configuration
gradient A defined by (2.5). Then the first Piola-Kirchhoff stress tensor P is given by
W W
P= , or Pij = , (3.4)
A Aij
where Pij and Aij are the (i, j) components of P and A, respectively. By (2.6) we can
rewrite (3.4) in terms of the Lagrangian strain tensor E,
W W
P=A , or Pij = Aik . (3.5)
E Ekj
By (2.13) and (3.5) we find that the second Piola-Kirchhoff stress tensor S is given by
W W
S= , or Sij = , (3.6)
E Eij
where Sij is the (i, j) component of S. By (2.12) and (3.6) we find that the Cauchy stress
tensor is given by
1 W T
= A A .
| A| E
Strain energy function for isotropic elastic materials

For an isotropic material, the configuration gradient A can be expressed uniquely in
terms of the principal stretches (i , i = 1, 2, 3) or in terms of the invariants (I1 , I2 , I3 )
of the left Cauchy-Green configuration tensor or right Cauchy-Green configuration
tensor (see Remark 2.3). Hence, we can express the strain energy function in terms of
principal stretches or in terms of invariants. Note that
1 = 2 = 3 = 1, I1 = 3, I2 = 3, and I3 = 1,
in the initial configuration where we choose W = 0. Thus a general formula for the
strain energy function can be expressed as
W ( 1 , 2 , 3 )
{[ ] }
= aijk 1i (2 + 3 ) + 2i (3 + 1 ) + 3i (1 + 2 ) (1 2 3 )k 6 ,
j j j j j j
(3.7a)
i,j,k=0
or

W ( I1 , I2 , I3 ) = cijk ( I1 3)i ( I2 3) j ( I3 1)k . (3.7b)
i,j,k =0
Due to their ubiquitous approximation properties, polynomial terms are usually cho-
sen in formulating strain energy functions, but the final forms are typically based on
empirical observations and are material specific for the choice of coefficients and trun-
cations. For incompressible materials (many rubber or elastomeric materials are often
nearly incompressible), | A| = 1 (which implies that 1 2 3 = 1 and I3 = 1), so (3.7a)
can be reduced to
[ ]
W (1 , 2 , 3 ) = aij 1i (2 + 3 ) + 2i (3 + 1 ) + 3i (1 + 2 ) 6 ,
j j j j j j
(3.8)
i,j=0
subject to
1 2 3 = 1,
and (3.7b) can be reduced to

W ( I1 , I2 ) = ci,j ( I1 3)i ( I2 3) j . (3.9)
i,j=0
Special cases for (3.9) include several materials:

1. A Neo-Hookean material for which W ( I1 , I2 ) = c10 ( I1 3),

2. A Mooney-Rivilin (or Mooney) material for which
W ( I1 , I2 ) = c10 ( I1 3) + c01 ( I2 3).
The Neo-Hookean and Mooney-Rivilin strain energy functions have played an im-
portant part in the development of nonlinear elasticity theory and its application. The
interested reader should consult [24, 44, 48, 59] and the references therein for further
information on hyperelastic materials.
3.2 Viscoelastic materials

The distinction between nonlinear elastic and viscoelastic materials is not always eas-
ily discerned and definitions vary. However it is generally agreed that viscoelasticity
is the property of materials that exhibit both viscous (dashpot-like) and elastic (spring-
like) characteristics when undergoing deformation. Food, synthetic polymers, wood,
soil and biological soft tissue as well as metals at high temperature display signifi-
cant viscoelastic effects. Throughout this section, we discuss the concept in a one-
dimensional formulation, such as that which occurs in the case of elongation of a sim-
ple uniform rod. In more general deformations one must use tensor analogues (as
embodied in (3.2)) of the stress, the strain and parameters such as modulus of elastic-
ity and damping coefficient.
In this section we first (Section 3.2.1) introduce some important properties of vis-
coelastic materials and then discuss the standard dynamic mechanical test in Section
3.2.2. We then present and discuss a number of specific forms of constitutive equations
proposed in the literature for linear viscoelastic materials (Section 3.2.3) and those for
nonlinear viscoelastic materials (Section 3.2.4).
3.2.1 Properties of viscoelastic materials

Viscoelastic materials are those for which the relationship between stress and strain
depends on time, and they possess the following three important properties: stress
relaxation (a step constant strain results in decreasing stress), creep (a step constant
stress results in increasing strain), and hysteresis (a stress-strain phase lag).
Stress relaxation
In a stress relaxation test, a constant strain 0 acts as input to the material from time
t0 , the resulting time-dependent stress is decreasing until a plateau is reached at some
later time, which is as depicted in Fig. 4. The stress function G (t) resulting from the
unit step strain (that is, 0 = 1) is referred to as the relaxation modulus.
In a stress relaxation test, viscoelastic solids gradually relax and reach an equilib-
rium stress greater than zero, i.e.,
lim G (t) = G > 0,

t
Figure 4: Stress and strain histories in the stress relaxation test.
Figure 5: Stress and strain histories in the creep test.
while for viscoelastic fluids the stress vanishes to zero, i.e.,
lim G (t) = 0.
t
Creep
In a creep test, a constant stress 0 acts as input to the material from time t0 , the
resulting time-dependent strain is increasing as depicted in Fig. 5.
The strain function J (t) resulting from the unit step stress (i.e., 0 = 1) is called the
creep compliance.
In a creep test, the resulting strain for viscoelastic solids increases until it reaches a
nonzero equilibrium value, i.e.,
lim J (t) = J > 0,

t
while for viscoelastic fluids the resulting strain increases without bound as t increases.
Hysteresis
Hysteresis can be seen from the stress-strain curve which reveals that for a viscoelastic
material the loading process is different than in the unloading process. For example,
the left plot in Fig. 6 illustrates the associated stress-strain curve for the Hookean elas-
tic solid, and that in the right plot of Fig. 6 is for the Kelvin-Voigt model (a linear
viscoelastic model discussed below in Section 3.2.3). From this figure, we see that we
can differentiate between the loading and unloading for the Kelvin-Voigt material, but
we cannot do this for Hookean elastic material. Thus the Kelvin-Voigt material re-
members whether it is being loaded or unloaded, hence exhibiting hysteresis in
the material.
(a) (b)
Figure 6: Stress and strain curves during cyclic loading-unloading. (a): Hookean elastic solid; (b) Kelvin-
Voigt material depicted by the solid line.
3.2.2 Dynamic mechanical tests: stress-strain phase lag, energy loss and complex
dynamic modulus
In addition to the creep and stress relaxation tests, a dynamic test is useful in studying
the behavior of viscoelastic materials. Stress (or strain) resulting from a small strain
(or stress) is measured and can be used to find the complex dynamic modulus as in-
troduced below. We illustrate these ideas with a discussion of the stress resulting from
a sinusoidal strain (as the discussion in the strain resulting from an analogous stress
can proceed similarly by just interchanging the role of stress and strain).
In a typical dynamic test carried out at a constant temperature, one programs a
loading machine to prescribe a cyclic history of strain to a sample rod given by
(t) = 0 sin(t), (3.10)
where 0 is the amplitude (assumed to be small), and is the angular frequency. The
response of stress as a function of time t depends on the characteristics of the material
which can be separated into several categories:
A purely elastic solid.
For this material, stress is proportional to the strain, i.e., (t) = (t). Hence with
the strain defined in (3.10), the stress is given by
(t) = 0 sin(t).
We find the stress amplitude 0 is linear in the strain amplitude 0 : 0 = 0 . The
response of stress caused by strain is immediate. That is, the stress is in phase with the
strain.
A purely viscous material.
For this kind of material, stress is proportional to the strain rate: (t) = d/dt.
For the strain defined in (3.10), the stress is then given by
(t) = 0 cos(t).
Note that ( )
cos(t) = sin t + ,
2
and thus we can rewrite the above expression as
( )
(t) = 0 sin t + .
2
The stress amplitude is linear in the strain amplitude: 0 = 0 , which is dependent
on the frequency . The stress is out of phase with the strain, and strain lags stress by
a 90 degree phase lag.
A linear viscoelastic solid.
With the sinusoidal strain (3.10), the stress as a function of time appears compli-
cated in the first few cycles. But a steady state will eventually be reached in which
the resulting stress is also sinusoidal, having the same angular frequency but re-
tarded in phase by an angle . This is true even if the stress rather than the strain is
the controlled variable. The cyclic stress is written as
(t) = 0 sin(t + ), (3.11)
where the phase shift is between 0 and /2, and the stress amplitude 0 depends on
the frequency . By an identity of trigonometry, we can rewrite (3.11) as
(t) = 0 cos() sin(t) + 0 sin() cos(t). (3.12)
Thus the stress is the sum of an in-phase response and out-of-phase response.
We consider the energy loss for a linear viscoelastic material such as described by
the stress (3.11) in response to the strain input (3.10). Let l be the length of a rod with
cross sectional area a. When the solid is strained sinusoidally, according to (3.10), the
solid elongates as
l (t) = l 0 sin(t).
By (3.12), the force on the rod is
F (t) = a (t) = a0 cos() sin(t) + a0 sin() cos(t).
During a time interval dt, the solid elongates by dl, and the work done on the rod is
dl
F (t)dl = F (t) dt = l 0 F (t) cos(t)dt.
dt
In one full cycle the work done is
2 2

W = al 0 0 cos() sin(t) cos(t)dt + al 0 0 sin() cos(t) cos(t)dt
0 0
=al 0 0 sin().
Note that the in-phase components produce no net work when integrated over a cycle,
while the out-of phase components result in a net dissipation per cycle equal to:
W = al 0 0 sin().
Thus, for a purely elastic solid, the stress is in phase with the strain ( = 0) and no
energy is dissipated. On the other hand, motion in the viscoelastic solid produces
energy loss.
It is a common practice in engineering to use complex variables to describe the
sinusoidal response of viscoelastic materials. Thus, instead of strain history (3.10), we
specify the complex strain as
= 0 exp(it).
Then we obtain the following complex stress instead of stress described by (3.11)
( )
= 0 exp i (t + ) .
The above equation can be rewritten as
= G ,
where G is defined by
0 0 0
G = exp(i) = cos() + i sin(). (3.13)
0 0 0
The characteristic parameter G is referred to as the complex dynamic modulus. We

denote the real part of G by G and the imaginary part of G by G. That is,
G = G + iG, (3.14)
where
0 0
G = cos(), and G = sin().
0 0
The coefficient G is called the storage modulus (a measure of energy stored and recov-
ered per cycle) which corresponds to the in-phase response, and G is the loss modulus
(a characterization of the energy dissipated in the material by internal damping) corre-
sponding to the out-of phase response. The in-phase stress and strain results in elastic
energy, which is completely recoverable. The /2 out-of-phase stress and strain re-
sults in the dissipated energy.
Remark 3.3. The relationship between the two transient functions, relaxation modulus
G (t) and creep compliance J (t), for a viscoelastic material is given by
t
J (s) G (t s)ds = t.
0
The relationship between the relaxation modulus G (t) and dynamic modulus func-
tions G and G is given by
( )
G ( ) = G + G (t) G sin(t)dt,
( 0 )
G( ) = G (t) G cos(t)dt.
0
The relations may be inverted to obtain

2 G ( ) G
G (t) = G + sin(t)d,
0
or

2 G( )
G (t) = G + cos(t)d.
0
The interested reader can refer to the recent text [33] for further information.
3.2.3 Linear viscoelastic models: constitutive relationships

The characteristic feature of linear viscoelastic materials is that the stress is linearly
proportional to the strain history, and it is important to note that the property of lin-
earity of response does not refer to the shape of any material response curve. Linear
viscoelasticity is usually applicable only for small deformations and/or linear mate-
rials. Thus, infinitesimal strain theory should be employed for this case. There are
two standard approaches that have been used to develop constitutive equations for
the linear viscoelastic materials: mechanical analogs and the Boltzmann superposi-
tion principle.
Mechanical analogs
Linear viscoelastic behavior can be conceived as a linear combinations of springs (the
elastic component) and dashpots (the viscous component) as depicted by Fig. 7. The
elastic component is described by
= ,
or
d 1 d
= ,
dt dt
where is the stress, is the strain that occurs under the given stress, and is the
elastic modulus of the material with units N/m2 . The viscous component is modeled
by
d
= ,
dt
(a) (b)
Figure 7: (a): Schematic representation of the Hookean spring; (b) Schematic representation of Newtonian
dashpot.
Figure 8: Schematic representation of the Maxwell model.
where is the viscosity of the material with units N s/m2 . The mechanical analogues
approach results in a linear ordinary differential equation with constant coefficients
relating stress and its rates of finite order with strain and its rates of the form
d d2 dn d d2 dn
a0 + a1 + a2 2 + + an n = b0 + b1 + b2 2 + + bn n . (3.15)
dt dt dt dt dt dt
In (3.15) the constant coefficients ai , bi , i = 0, 1, 2, , n are related to the elastic mod-
ulus and viscosity of the material which are usually determined from physical ex-
periments. A complete statement of the constitutive equation obtained from use of
mechanical analogs then consists of both an equation of the form (3.15) and a set of
appropriate initial conditions. In addition, we see that it is not convenient to directly
incorporate this general model (3.15) into the equation of motion.
The three basic models that are typically used to model linear viscoelastic materials
are the Maxwell model, the Kelvin-Voigt model and the standard linear solid model.
Each of these models differs in the arrangement of these springs and dashpots.
The Maxwell model

The Maxwell model is represented by a purely viscous damper and a purely elastic
spring connected in series as depicted in Fig. 8.
Because the spring and the dashpot are subject to the same stress, the model is also
known as an iso-stress model. The total strain rate is the sum of the elastic and the
viscous strain contributions, so that
1 d d
+ = . (3.16)
dt dt
We consider the stress relaxation function and creep function for the Maxwell model
(3.16). The stress relaxation function corresponds to the relaxation that occurs under
an imposed constant strain, given
( t ) = 0 H ( t t0 ), and (0) = 0,
where H (t) is the Heaviside step function (also called the unit step function in the liter-
ature), t0 0. The solution (t) to (3.16) is the relaxation function. With this strain
function, (3.16) can be written as
1 d
+ = 0 ( t t0 ),
dt
where is the Dirac delta function. Let
(s) = L { (t)}(s),
where L denotes the Laplace transform. Then taking the Laplace transform of both
sides of the above differential equation we obtain
(1 1 ) 1 ( ) 1
(s) = 0 et0 s + s = 0 et0 s s + .

Taking the inverse Laplace transform one finds that

[ ]
(t) = exp (t t0 ) 0 H (t t0 ).

The stress relaxation function for the Maxwell model (3.16) is illustrated in Fig. 9 (com-
pare with Fig. 4).
The creep function corresponds to the creep that occurs under the imposition of a
constant stress given by
(t) = 0 H (t t0 ), and (0) = 0,
the solution (t) to (3.16) is the creep function. With this stress function, (3.16) can be
written as
d 0 0
= H ( t t0 ) + ( t t0 ).
dt
Then taking the Laplace transform of both sides of the above differential equation we
have that
0 et0 s 0 et0 s
(s) = + .
s2 s
Upon taking the inverse Laplace transform we find

[1 1 ]
(t) = + (t t0 ) 0 H (t t0 ).

The creep function of Maxwell model (3.16) is illustrated in Fig. 10 (again, compare
with Fig. 5).
Figure 9: Stress relaxation function of Maxwell model.
Figure 10: Creep function of Maxwell model.
From the above considerations we see that the Maxwell model predicts that stress
decays exponentially with time, which is accurate for many materials, especially poly-
mers. However, a serious limitation of this model (with creep as depicted in Fig. 10) is
its inability to correctly represent the creep response of solid material which does not
increase without bound. Indeed polymers frequently exhibit decreasing strain rate
with increasing time.
Finally we find the storage modulus G and loss modulus G for the Maxwell
model. Let ( )
(t) = 0 exp(it), and (t) = 0 exp i (t + ) .
Then we substitute and into (3.16) which after some algebraic arguments results in
the complex dynamic modulus
0 ( )2 2
G = exp(i) = 2 + i .
0 + ( )2 2 + ( )2
For the storage and loss modulus we thus find
( )2 2
G = , G = .
2 + ( )2 2 + ( )2
By taking the derivative of G with respect to frequency , we find that the loss mod-
ulus achieves its maximum value at = 1/, where = / is the relaxation time.
The Kelvin-Voigt model

The Kelvin-Voigt model, also known as the Voigt model, consists of a Newtonian
damper and a Hookean elastic spring connected in parallel, as depicted in Fig. 11.
Figure 11: Schematic representation of the Kelvin-Voigt model.
Because the two elements are subject to the same strain, the model is also known as an
iso-strain model. The total stress is the sum of the stress in the spring and the stress in
the dashpot, so that
d
= + . (3.17)
dt
We first consider the stress relaxation function and creep function for the Kelvin-Voigt
model (3.17). The stress relaxation function corresponds to the solution of (3.17) when
( t ) = 0 H ( t t0 ), and (0) = 0.
We find
(t) = 0 H (t t0 ) + 0 (t t0 ).
This stress relaxation function for the Kelvin-Voigt model (3.17) is illustrated in Fig. 12.
The creep function again is the solution (t) to (3.17) corresponding to (t) =
0 H (t t0 ) and (0) = 0. We find
d
+ = 0 H (t t0 ).
dt
Let (s) = L {(t)}(s), or in terms of the Laplace transform we have
e t0 s 0 [ et0 s e t0 s ]
(s) = 0 = .
s( + s) s s + /
Using the inverse Laplace transform we obtain
1[ ( )]
(t) = 1 exp (t t0 ) 0 H (t t0 ).

The corresponding creep function for the Kelvin-Voigt model (3.17) is illustrated in
Fig. 13 (compare with Fig. 5).
Thus we find that the Kelvin-Voigt model is extremely accurate in modelling creep
in many materials. However, the model has limitations in its ability to describe the
commonly observed relaxation of stress in numerous strained viscoelastic materials.
Figure 12: Stress relaxation function for the Kelvin-Voigt model.
Figure 13: Creep function of Kelvin-Voigt model.
Finally we may obtain the storage modulus G and loss modulus G for the Kelvin-
Voigt model (3.17). Letting
( )
(t) = 0 exp(it), and (t) = 0 exp i (t + ) ,
and substituting and into (3.17), we find

0
G = exp(i) = + i.
0
Hence, by (3.14) we have
G = , G = .
Remark 3.4. Because of one of our motivating applications (discussed in Section 4
below), we are particularly interested in the elastic/viscoelastic properties of soil.
Hardin in [31] presented an analytical study of the application of the Kelvin-Voigt
model to represent dry soils for comparison with test results. From this study he
found that the Kelvin-Voigt model satisfactorily represented the behavior of sands in
these small-amplitude vibration tests if the viscosity in the model was treated as
varying inversely with the frequency to maintain the ratio / constant. Since
Hardins work, describing soils as a Kelvin-Voigt material has become accepted as
one of the best ways in soil dynamics of calculating wave propagation and energy
dissipation. The Kelvin-Voigt model also governs the analysis of standard soil tests,
including consolidation and resonant-column tests. The interested reader can refer
to [40, 41, 47] as well as the references therein for more information of the models
historical backgrounds and examples of its application in soil dynamics.
It is interesting to note that the author in [13] showed that the Kelvin-Voigt model
can be used to describe the dynamic response of the saturated poroelastic materials
that obey the Biot theory (the two-phase formulation of Biot in [15]). This viscoelas-
tic model is simpler to use than poroelastic models but yields similar results for a
wide range of soils and dynamic loadings. In addition, the author in [40] developed a
model, the Kelvin-Voigt-Maxwell-Biot model, that splits the soil into two components
(pore fluid and solid frame) where these two masses are connected by a dashpot which
can then be related to permeability. In addition, a mapping between the Kelvin-Voigt
model and the Kelvin-Voigt-Maxwell-Biot model is developed in [40] so that one may
continue to use the Kelvin-Voigt model for saturated soil.
The standard linear solid (SLS) model
The standard linear solid model, also known as the Kelvin model or three-element
model, combines the Maxwell Model and a Hookean spring in parallel as depicted
in Fig. 14.
The stress-strain relationship is given by
d ( d )
+ = r + , (3.18)
dt dt
where
1 r + 1
= , and = 1 ,
1 r 1
from which can be observed that > .

The stress relaxation function and the creep function for the standard linear model
(3.18) are obtained in the usual manner. As usual, the stress relaxation function (t) is
obtained by solving (3.18) with (t) = 0 H (t t0 ) and (0) = 0. We find
d
(t) + = r 0 H (t t0 ) + r 0 (t t0 ),
dt
Figure 14: Schematic representation of the standard linear model.

or in terms of the Laplace transform
e t0 s e t0 s
(s) =r 0 + r 0
s(1 + s) 1 + s
e t 0 s ( ) e t0 s
= r 0 + r 0 1 .
s s + 1
Thus we find
( ) [ tt ]
0
( t ) = r 0 H ( t t0 ) + r 0 1 exp H ( t t0 )

[ ( ) ( t t )]
0
= r 1 + 1 exp 0 H ( t t0 )

[ ( t t )]
0
= r + 1 exp 0 H ( t t0 ).

This stress relaxation function for the standard linear model (3.18) is illustrated in
Fig. 15.
The creep function is the solution of (3.18) for (t) given (t) = 0 H (t t0 ) and
(0) = 0. Using the same arguments as above in finding the stress function, we have
1[ (

) ( t t )]
0
(t) = 1+ 1 exp 0 H (t t0 ).
r
The creep function of the standard linear model (3.18) is illustrated in Fig. 16.
We therefore see that the standard linear model is accurate in predicating both
creep and relaxation responses for many materials of interest.
Figure 15: Stress relaxation function for the standard linear model.
Figure 16: Creep function for the standard linear model.

Finally the usual arguments for the standard linear model lead to
0 r (1 + 2 ) r ( )
G = exp(i) = +i .
0 1 + ( )2 1 + ( )2
By definition of the storage and loss modulus, we thus have
r (1 + 2 ) r ( )
G = , G = .
1 + ( )2 1 + ( )2
Remark 3.5. It was demonstrated in [45] that the relaxation behavior of compressed
wood can be adequately described by the standard-linear-model.
The Boltzmann superposition model

A general approach widely used to model linear viscoelastic materials is due to Boltz-
mann (1844-1906) and is called the Boltzmann superposition model or simply the
Boltzmann model.
If the origin for time is taken at the beginning of motion and loading (i.e., (0) = 0
and (0) = 0), then the stress-strain law is given by
t
d(s)
(t ) = r (t) + K (t s) ds, (3.19)
0 ds
where r represents an instantaneous relaxation modulus, and K is the gradual re-
laxation modulus function. The relaxation modulus function G (t) for the Boltzmann
model (3.19) is given by
G ( t ) = r + K ( t ).
Note that (0) = 0. Hence, (3.19) can be rewritten as follows
t
d(s)
(t) = G (t s) ds. (3.20)
0 ds
If the strain (t) has a step discontinuity at t = 0, then by integration of the resulting
delta function for its derivative we obtain the following representation from (3.20)
t
d(s)
( t ) = G ( t ) (0) + G (t s) ds. (3.21)
0 ds
This results in a decaying stress if the strain is held constant after a step discontinuity.
The interested reader can refer to [16, pp. 56] for more information on the connection
of several different forms of the Boltzmann superposition model.
We find that when model (3.19) is incorporated into force balance laws (equation
of motion), it results in integro-partial differential equations which are most often phe-
nomenological in nature as well as being computationally challenging both in simula-
tion and control design.
Remark 3.6. Any constitutive equation of form (3.15), along with the appropriate ini-
tial conditions, can be expressed in either the form (3.19) (or (3.21)) or its correspond-
ing compliance form (i.e., the strain in terms of the stress and stress rate). On the other
hand, a constitutive equation of form (3.19) (or (3.21)) can be reduced to the form (3.15)
if and only if the stress relaxation modulus satisfy specific conditions. A detailed dis-
cussion of this can be found in [28]. For example, the Maxwell, Kelvin-Voigt, and
standard linear models can be expressed in the Boltzmann formulation. The choice of
parameters in (3.19) to yield these models are readily verified to be:
[ ]
1. The Maxwell model: r = 0, and K (t s) = exp (t s)/ with = /;
2. The Kelvin-Voigt model: r = , and K (t s) = (t s);
( ) [ ]
3. The standard linear model: K (t s) = / 1 r exp (t s)/ , which can be
[ ]
simplified as K (t s) = 1 exp (t s)/ with = 1 /1 .
Internal variable approach

Other special cases of (3.19) that are often encountered include a generalization of
the single spring-dashpot paradigm of Fig. 14 (the standard linear model) to one with
multiple spring-dashpot systems in parallel as depicted in Fig. 17. This is a generalized
standard linear model (also referred to as generalized Maxwell model or Wiechert model or
Maxwell-Wiechert model in the literature) that results in the relaxation response kernel
n ( t s)
K (t s) = j exp
j
, (3.22)
j =1
where j = j / j , j = 1, 2, , n. We observe that the Boltzmann formulation with

relaxation response kernel (3.22) is equivalent to the formulation
n
( t ) = r ( t ) + j j ( t ), (3.23)
j =1
where the j satisfy the following ordinary differential equations
d j (t) 1 d(t)
+ j (t) = , j (0) = 0, j = 1, 2, , n. (3.24)
dt j dt
The formulation (3.23) with (3.24) is sometimes referred to as an internal variable model
(e.g., see [7,12]for other discussions of internal state variable approaches, see also [35,
36, 5153]) because the variables j can be thought of as internal strains that are
driven by the instantaneous strain according to (3.24) and that contribute to the total
stress via (3.23). Hence, we can see that this internal variable modeling leads to an
efficient computational alternative for the corresponding integro-partial differential
equation models involving (3.19). In addition, we note that this approach provides a
molecular basis for the models as it can be thought of as a model for a heteroge-
neous material containing multiple types of molecules [7, 12, 19, 32], each possessing a
Figure 17: Schematic representation of the generalized standard linear model.
distinct relaxation property characterized by a relaxation parameter j . For a compari-

son of models of viscoelastic damping via hysteretic integrals versus internal variable
representations, see [7] and the references therein.
Moreover, we see that model (3.23) with (3.24) can be readily viewed as a special
case of or even an approximation to models with a continuum of relaxation times
(see [12] and the references therein). These models have proved useful in a wide range
of viscoelastic materials. The corresponding stress-strain laws have the form

d(t)
(t; P ) = r (t) + + 1 (t; )dP ( ), (3.25)
dt T
where P is a probability distribution over the set T of possible relaxation parameters

, and 1 (t; ) satisfies, for each T ,
d 1 (t; ) 1 d(t)
+ 1 (t; ) = , 1 (0; ) = 0. (3.26)
dt dt
The approach embodied in equations (3.25)-(3.26) offers a computationally tractable

alternative for linear materials with a continuum of relaxation times.
Remark 3.7. The generalized standard linear model (i.e., model (3.23) combined with
(3.24)) with different n has been successfully used to describe the stress relaxation be-
havior of a variety of foods. For example, in [57] this model with n = 2 was success-
fully used to describe the stress relaxation of lipids such as beeswax, candelilla wax,
carnauba wax and a high melting point milkfat fraction. A comprehensive study on
the ability of the generalized Maxwell model to describe the stress relaxation behav-
ior of solid food is presented in [18]. In this study, five different food matrices (agar
gel, meat, ripened cheese, mozzarella cheese and white pan bread) were chosen as
representatives of a wide range of foods, and results verify that the proposed model
satisfactorily fits the experimental data.
3.2.4 Nonlinear viscoelasticity: constitutive relationships

For many materials the linear models given in Section 3.2.3 are inadequate to de-
scribe experimental data. In particular, shape memory alloys such as Nitinol (a nickel-
titanium) and CuZnAl (a cooper zinc aluminum aloy), biological soft tissue and highly
filled rubber exhibit significant nonlinear hysteric behavior such as that depicted by
the experimental data for highly filled rubber given in Fig. 18. Nonlinear viscoelastic
behavior is usually exhibited when the deformation is large or if the material changes
its properties under deformations.
Figure 18: Experimental stress-strain curves for (1) unfilled, (2) lightly filled and (3) highly filled rubber in
tensile deformations.
The theory of nonlinear viscoelasticity has attracted the attention of a large num-
ber of investigators over the past century (e.g., [19, 20, 23, 25, 37, 42, 53, 58]). Two types
of models for stress-strain relationships can be found in the literature. One is based
on the phenomenological mechanical behavior of the materials (that is, the form of
constitutive equations is not based on the explanation of how these properties arise
from the underlying microscopic structure). For example, Green and Rivlin in [27]
constructed a multiple integral constitutive equation, which is arranged as a series
in which the nth term is of degree n in the strain components. A multiple integral
constitutive equation, arranged in a series, was also developed by Pipkin and Rogers
in [46], in which the first term gives the results of a one-step test (the stress due to
a step change of strain), and whose nth term represents a correction due to the nth
step. The interested reader can refer to [20, 61] for recent historical overviews on these
phenomenological models as well as the mathematical issues underlying the formu-
lations of these models. The other type of model entails formulations based on the
molecular mechanisms underlying the response. For example, in [19] Doi and Ed-
wards developed a reptation model for concentrated solutions and polymer melts
which is based on the assumption that an entangled polymer molecule, the chain,
slides (or reptates) through a tube whose contours are defined by the locus of en-
tanglements with neighboring molecules. The polymer chain is free to diffuse along
the tube axis but cannot move perpendicularly to the tube as other molecules restrict
such movement. In [14] the polymer melt is modeled as a collection of interacting
bead-rod chains. This model, referred to as the Curtiss-Bird model, is similar to the
Doi-Edwards model, but it is based on a systematic kinetic development and does not
use the phenomenological constraints of a chain in a tube. In addition, the compar-
ison results of the Curtiss-Bird model and the Doi-Edwards model in [14] show that
the the Curtiss-Bird model is more accurate than Doi-Edwards model in predicting the
nonlinear behavior. A review of these molecular types of models as well as the rela-
tionship between the nonlinear viscoelasticity and molecular structure is given in [43].
In the following sections we restrict our discussions to those models that have
been employed and developed by our group for understanding the dynamic response
of the highly filled rubber and propagation of arterial stenosis induced shear waves in
composite biotissue. The forms of these constitutive equations can be easily incorpo-
rated into the equation of motion so that we can numerically solve a partial differential
equation with appropriate boundary and initial conditions to understand the dynamic
behavior of materials.
Modified Boltzmann superposition model

The most direct formulation to treat nonlinear viscoelasticity is one based on general-
izing the Boltzmann superposition model (3.19) to a corresponding nonlinear version.
That is, one allows nonlinear instantaneous strain as well as nonlinear strain rate de-
pendence. This modified superposition principle was first suggested in [34] where
it was observed the creep behavior of fibres could be separated into time and stress-
dependent parts, and then considered in [22] for a multiple-step test (in this regard
see also the numerous contributions of Schapery as discussed in [5153]). One form
of this approach has been employed in some of our earlier efforts [9, 10] for modeling
hysteretic damping in elastomers and is given by
t
d ( )
(t) = ge ((t)) + c D (t) + K (t s) gv (s), (s) ds, (3.27)
0 ds
where is the infinitesimal strain, K is the convolution memory kernel, and ge and
gv are nonlinear functions accounting for the elastic and viscoelastic responses of the
materials, respectively. As explained in [9], our nonlinear materials undergoing large
deformations required the use of finite (as opposed to infinitesimal) strain theories.
However, since the nonlinearity between the stress and finite strain is an unknown and
is to be estimated (using inverse problem algorithms) and since the finite strain can be
expressed in terms of known nonlinearities as a function of the infinitesimal strain (at
least in the problems of interest here), one can effectively formulate the problem as
one of estimating the unknown nonlinearity between stress and infinitesimal strain.
Thus one can develop nonlinear models for large deformation-related stress in terms
of infinitesimal strain.
Again, this form of model (3.27), when incorporated into force balance laws (equa-
tion of motion), results in integro-partial differential equations which are computa-
tionally challenging both in simulation and control design. This motives us to employ
the internal variable approach described below as an alternative for computation.
Fungs quasi-linear viscoelastic formulation

This formulation can be found in Fung [25] in the context of efforts to model soft tissue,
which is often characterized by a constant hysteresis over a wider frequency range. In
this formulation, a continuous spectrum over a finite range (1 , 2 ) is used to describe
this phenomenon instead of using a finite spectrum {1 , 2 , , n } associated with
relaxation kernels of the form given in (3.22). Fungs model is given by
t
de (( ))
(t) = K (t ) d. (3.28)
0 d
Here the stress is the force in the deformed configuration divided by the cross-
sectional area of the speciman at the zero stress state (i.e., the scalar version of the first
Piola-Kirchhoff stress tensor), is the ratio of the length of the specimen stretched
under the load divided by the initial length at the zero stress state, e is the elastic
response to this elongation, and K (t) is the reduced relaxation function (reduced
refers to that K (t) is normalized such that K (0) = 1). Fung proposes a kernel of the
form ( )
1 + c 12 1 exp t d
K (t) = , (3.29)
1 + c ln(2 /1 )
where c represents the degree to which viscous effects are present, and 1 and 2 rep-
resent fast and slow viscous time phenomena.
Since its introduction, this quasi-linear viscoelastic (QLV) theory of Fung has been
applied successfully in stress-strain experiments to several types of biological tissue.
A benefit to using (3.28) as a constitutive equation is that, unlike simpler models for
viscoelasticity, it allows for the consideration of a continuous spectrum (e.g., see the
discussions in [25]) of relaxation times; this is also true of the probabilistic-based in-
ternal variable approach developed in [8] and described below. The need for a con-
tinuum of relaxation times in certain materials was observed many years ago (e.g.,
see [21, 29, 55]). While Fungs theory has been successfully employed for fitting hys-
teretic stress-strain curves, for control applications one is interested in using it in a full
dynamical model. Unfortunately, the QLV theory, as presented by Fung, leads to ex-
ceedingly difficult computations within full dynamical partial differential equations,
especially in estimation and control problems.
Internal variable approach

To overcome the computational challenges present in the modified Boltzmann super-
position model and Fungs quasi-linear viscoelastic model, we turn to an internal vari-
able approach in an alternative way to formulate the constitutive equation.
We observe that if the relaxation response kernel function K in model (3.27) is de-
fined as in (3.22), then (3.27) can be written in terms of internal strains similar to (3.23)
with (3.24):
n
(t) = ge ((t)) + c D (t) + j j (t), (3.30)
j =1
where the internal strains j satisfies

d j (t) 1 ( )
+ j (t) = gv (t), (t) , j (0) = 0, j = 1, 2, , n. (3.31)
dt j
Through comparison with experimental data, in [10] it was discovered that the best
fit to filled elastomer data occurs when ge and gv are cubic. Moreover, model (3.30)
with (3.31) can be readily generalized to models with a continuum of relaxation times
(see [12] and the references therein). These models have been useful in a wide range
of viscoelastic materials. The corresponding stress-strain laws have the form

(t; P ) = ge ((t)) + c D (t) + 1 (t; )dP ( ), (3.32)
T
where P is a probability distribution over the set T of possible relaxation parameters

, and 1 (t; ) satisfies, for each T ,
d 1 (t; ) 1
+ 1 (t; ) = gv ((t), (t)), 1 (0; ) = 0. (3.33)
dt
We also observe that if the reduced relaxation function K (t) in Fungs QLV model
(3.28) is defined as in (3.22) with
n
j = 1,
k =1
then model (3.28) is equivalent to the following internal strain variable formulation
n
(t) = j j ( t ), (3.34)
j =1
with the internal strains j satisfying

d j (t) 1 de ((t))
+ j (t) = , j (0) = 0, j = 1, 2, , n. (3.35)
dt j dt
Various internal strain variable models are investigated in [1] and a good agreement
is demonstrated between a two internal strain variable model (i.e., n = 2) and un-
damped simulated data based on the Fung kernel (3.29). The corresponding stress-
strain laws for a generalization of these models to have a continuum of relaxation
times have the form

(t; P ) = 1 (t; )dP ( ), (3.36)
T
where once again P is a probability distribution over the set T of possible relaxation
parameters , and 1 (t; ) satisfies, for each T ,
d 1 (t; ) 1 de ((t))
+ 1 (t; ) = , 1 (0; ) = 0. (3.37)
dt dt
Molecular type of constitutive equations

Even though model (3.30) combined with (3.31) provides a reasonable fit to the ex-
perimental data, it does not provide insight into the underlying mechanisms for ten-
sile and/or shear deformations in filled rubber. This is not unexpected since model
(3.30)-(3.31) is based on pseudo-phenomenological formulations. Hence, a different
approach based on molecular arguments was pursued in [46], where the ideas of
these models are based on those of Johnson and Stacer in [32]. It turns out that this
approach leads precisely to a class of models based on a Boltzmann formulation.
A polymer material undergoing directional deformation is modeled in [2] in which
polymer chains are treated as Rouse type strings of beads interconnected by springs
(see [49]) as depicted in the left plot of Fig. 19. The model permits the incorporation of
many important physical parameters (such as temperature, segment bond length, in-
ternal friction, and segment density) in the overall hysteretic constitutive relationship.
The model in [2] was based on the assumption that the materials were composed of
two virtual compartments as depicted in the right plot of Fig. 19. One compartment
consists of a constraining tube which is a macroscopic compartment containing both
CC (chemically cross-linked) and PC (physically constrained) molecules. The other
compartment is microscopic in nature and consist of those PC molecules aligned with
the direction of the deformation. These molecules will at first stick to the constrain-
ing tube and be carried along with its motion, but will very quickly slip and begin
to relax back to a configuration of lower strain energy. In the model derivation one
computes the contributions of both compartments to the overall stress of this poly-
(a) (b)
Figure 19: (a): Representation of vectors for a bead-spring polymer molecule; (b): PC molecule entrapped
by the surrounding constraining tube.
mer material undergoing deformations to obtain the constitutive law, which is similar
to that developed in [4, 5] and has the general form of Boltzmann type model (3.27),
even though the kernel is not of convolution type.
4 Examples
We have introduced and briefly discussed a number of possible constitutive relation-
ships that represent a noncomprehensive review of an extensive body of research lit-
erature on elastic/viscoelastic materials. We conclude by presenting an example to
illustrate how one might apply the equations of motion and a constitutive relation-
ship to model a particular situation. In our computational example, we will examine
the properties of a one dimensional column of dry soil and study the movement of the
soil in response to a sinusoidal input at the surface. At first, we will record displace-
ments in a wave moving past a stationary observation point in the soil column. We
then add a rigid body to the column to demonstrate the changes in wave propagation
one might expect if something is buried in the soil. We can change parameters, such
as soil density, and understand the impact these changes have on the displacements
at the observation point. In all cases, we focus on seismic P waves (longitudinal
waves) propagating downward through the soil away from the source of the force (an
impact) located at the ground surface.
4.1 Model description

The schematic for this problem is given on the left side of Fig. 20 where the wave ob-
servation point is at the location z = z10 and the ground surface (as well as the source
of the sinusoidal input force) is at z = z p0 . The right side of the figure represents the
idealized (1-dimensional) mathematical soil column with input source and observa-
tion point marked in the figure. In the case where we have a rigid object in the soil,
that object is placed at z = z10 in the column instead of an observation point.
In this idealized configuration, we assume that both soil and target are uniform
in x- and y-directions. Based on the discussion in Remark 3.4, we assume that soil
air
.
soil
soil
Figure 20: (left): Schematic of problem; (right): 1-dimensional representation.

behaves as a Kelvin-Voigt material for small vibrations. This means that stress com-
ponents in soil can be expressed as sum of two terms, the first term being proportional
to the strain () and the second term being proportional to the rate of change () of
strain.
Let u(z, t) denote the displacement (in units m) in the z-direction at position z at
time t. Then in the situation with no buried object, for z (z p0 , ) we have
Case 1: Only soil present in the column
2 u(z, t) ( u(z, t) 2 u(z, t) )

(z) = ( z ) + ( z ) . (4.1)
t2 z z tz
For a second case, we assume there is a buried rigid object with its center of mass
located at z = z10 . We assume that the target is homogeneous in the z-direction, and
the contacting surface area between the target and the soil under the target is the same
as the one between the target and the soil above the target. Since the target is a rigid
body, the displacement of upper side of the target is exactly the same as that of its
lower side. Hence, in the 1-dimensional setting we can treat the target as a point mass.
With the given assumptions on soil and target we can visualize the soil-target-soil as
two thin rods connected by a point mass at z = z10 . The schematic of the problem is
illustrated in the right plot of Fig. 20, where in this case there is a point mass at z = z10 .
The resulting pair of equations are
Case 2: A rigid target present in the soil column
2 u(z, t) ( u(z, t) 2 u(z, t) )

(z) = ( z ) + ( z ) , z (z p0 , z10 ) (z10 , ), (4.2a)
t2 z z tz
( )
+ u ( z10 , t ) + u ( z10 , t )
+ +
2 u(z10 , t) 2
M = S ( z ) + ( z )
t2 10
z 10
tz
( u )
u ( z10 , t )
( z , t ) 2

S (z10 ) 10
+ (z10 ) . (4.2b)
z tz
In all cases denotes the density (in units kg/m3 ) of soil, is the elastic modulus (in
units kg/m s 2 = Pa) of soil , and represents the damping coefficient (in units kg/m
s) of soil. For the second equation in (4.2), M is used to denote the mass (in units
kg) of the target and S represents the surface area (in units m2 ) of contact between
the target and the soil under (or above) the target. Though the model will treat non-
constant and piecewise-defined coefficients, for our example we will take the simple
case where the parameters are constant values in the soil column. In general, non-
constant and piecewise-defined coefficients would allow one to account for varying
physical situations, such as having a lower density above a buried object than below
the buried object. For initial conditions we will assume zero displacement and zero
velocity since the system is at rest initially. These conditions are then given by
u
u(z, 0) = 0, (z, 0) = 0. (4.3)
t
The boundary condition at z = z p0 is given by

( u(z, t) 2 u(z, t) )
(z) + (z) = f ( t ), (4.4)
z tz z=z p0
where f is the applied external force in units N/m2 . On the surface, the normal inter-
nal stress is balanced with the input force, resulting in (4.4).
In order to numerically solve our model (4.1) with (4.3) and (4.4) or model (4.2)
with (4.3) and (4.4), it is convenient to have a finite spatial domain. Since we only care
about the displacements near the surface (and near the buried object in the second
case), we will choose the right (lower) boundary z00 to be sufficiently far away from
the target so that no energy will reach the boundary z00 during the time frame within
which we run the simulations. This assumption implies that we can set up any type
of boundary condition at z00 . For simplicity, we assume that
u(z00 , t) = 0. (4.5)
Hence, our problem (4.1) with (4.3) and (4.4), and problem (4.2)-(4.4) is thus defined
on the finite space domain [z p0 , z00 ]. We report on computations for the model using a
standard finite element method.
4.2 Simulation results

We report the results of some of our simulations with the equations governing this
one layer problem, observing the wave form through time at location z10 = 0.3048m,
which is approximately one foot beneath the ground surface at z p0 = 0. The value
for the far boundary was set at 50m, and no reflections from the far boundary were
observed in the calculations. The baseline numerical values for soil density and elastic
modulus are
= 1800kg/m3 , = 2.04 108 Pa, (4.6)
where the values were estimated from [17]. For the value of damping coefficient of soil
, we will assume that damping is frequency- and elastic modulus-dependant, via the
formula
2
= . (4.7)
1 2
The parameter is called the damping ratio, which is related to the energy lost be-
tween wave peaks. The baseline value for the damping ratio was set at = 0.05,
which again was derived using results from [17]. For the carrier frequency in our
simulations, we apply the sinusoidal input function at a frequency of = 400.
We depict a plot of a sinusoidal input function in Fig. 21. Recall that, given our
coordinate system, positive force values represent downward force and negative force
values represent upward force. Thus, this input represents impacting the ground in a
downward motion and then the ground rebounding with equal force. We can thus see
5
x 10
Forcing Function
4
f(t) (units:N/m )
2
1
4
0 0.002 0.004 0.006 0.008 0.01
t (units: s)
Figure 21: Sinusoidal input function.
displacement return to the baseline in all of the following figures due to the restoring
force present in the input. Since the displacement returns to the baseline, we can also
infer that our model is linearly dependent on the input.
We remark that the input term and the equations we use for demonstration pur-
poses are only a convenient approximation to physical reality in a small displacements
case. One could implement the actual physical situation by allowing for a moving up-
per boundary at z p0 . In this case, the input would only be the first, positive part of the
sine wave, as the differential equation dynamics coupled with the moving boundary
would return the soil to near its original position. Since a moving boundary is more
difficult to implement computationally, for this demonstration we chose to implement
the simpler stationary boundary at the impact site z p0 . In reality, the ground boundary
will not remain stationary under impact, but it will instead move first in the positive
(downward) direction and then rebound in the negative (upward) direction. The re-
bound is due in part to the viscoelastic properties of the soil column and also to the
physical soil column interacting with the surrounding soil (e.g., shear which is not
represented explicitly in the one dimensional dynamics). We model this restoring mo-
tion as the second half of the input signal as depicted in Fig. 21. That is, the positive
first half of the sinusoid represents the force imparted by the thumper and the nega-
tive latter half is modeling the rebounding of the soil boundary that one would (and
we did in field experiments) see in reality. This permits use of the stationary boundary
at z p0 while still approximating (with reasonable accuracy) the true dynamics at the
surface.
4.2.1 Results for Case 1: only soil present in the column

The first situations we examine are when holding one parameter (of (, )) constant
and increase the other parameter. The results are depicted in the panels of Fig. 22.
In the left panel, we hold density constant and change the elastic modulus . As
the modulus increases, we see that the wave begins its upward slope sooner, meaning
the wave has arrived at the observation point sooner so wave speed has increased.
Also, as the modulus increases, we see less displacement overall. In the case where
density increases, we see that we also get less displacement overall, but the wave
speed decreases. We clearly see the model demonstrates that soil parameters have
multiple realistic effects on overall displacement and wave speed.
In Fig. 23, we examine the situation where we increase both soil parameters. As ex-
pected, the effects of the parameter increases combine to reduce overall displacement.
In the left pane of the figure, we see that the elastic modulus increases more rapidly
than the density and the wave reaches the observation point sooner. In the right pane,
we applied the same percentage increases to the parameters and so the wave reaches
the observation point at the same time for all the parameter combinations. Thus, our
model is able to demonstrate the effects of more complex parameter changes that one
might see in applications. For example one might input multiple impacts at the same
site, which would increase both soil density and stiffness from one impact to the next.
(This in fact was realized in field experiments by scientific colleagues when testing for
repeatability of responses to interrogation impacts.)
4
x 10 Displacement around z=0.3048m (z=1 ft) 4
x 10 Displacement around z=0.3048m (z=1 ft)
12 10
=102000000 =1800 =204000000 =1440

10
=204000000 =1800 8 =204000000 =1800
u(0.3048,t)) (units: m)
=408000000 =1800
u(0.3048,t)) (units: m)
=204000000 =2250
8
6
6
4
4
2
2
0 0
2 2
0 0.002 0.004 0.006 0.008 0.01 0 0.002 0.004 0.006 0.008 0.01
t (units: s) t (units: s)
Figure 22: Wave form at z10 with varying soil parameters individually (left: variable , constant ; right:
constant , variable ).
4
x 10 Displacement around z=0.3048m (z=1 ft) 4
x 10 Displacement around z=0.3048m (z=1 ft)
14 12
=102000000 =1440 =163200000 =1440
12 =204000000 =1800 10
=204000000 =1800
=408000000 =2250 =255000000 =2250
u(0.3048,t)) (units: m)
u(0.3048,t)) (units: m)
10
8
8
6
6
4
4
2
2
0 0
2 2
0 0.002 0.004 0.006 0.008 0.01 0 0.002 0.004 0.006 0.008 0.01
t (units: s) t (units: s)
Figure 23: Wave form at z10 with both soil parameters and varying.
4.2.2 Results for Case 2: rigid body present in the soil column
In this section we examine the case where we have included a rigid body located at
z10 , modeled as a point mass at the same point as our previous observation point. In
the figures, this location is represented by a vertical dashed line. For all the simulation
results presented in this section, the values for the soil density, elastic modulus and
damping ratio are chosen to be the baseline values specified earlier, with the value for
the mass of the target given by M = 34.2671kg, and the contacting surface area taken
as S = 0.1580m2 .
In the upper left pane of Fig. 24, we can see that the sinusoidal force has begun
imparting displacement in the soil but this displacement has not yet reached the target.
In the upper right pane, the displacement has impacted the target. In this simulation,
the target imparts much of the energy to the soil below it. In the lower left pane of
Fig. 24, we see that most of the energy has passed through the target and is deeper into
the soil. However, looking at the domain between z = 0 and the dashed target line,
we see the remnant energy that was reflected by the target back toward the surface. In
the lower right pane in the figure, we see that this energy has bounced off the soil and
impacted the target again, once more passing some energy through the target (as seen
Wave form in zdomain, at time t=0.00067831 Wave form in zdomain, at time t=0.0020486
5 4
x 10 x 10
14
12 8
Displacement, units: m
10
6
8
6 4
4
2
2
0 0
2
0 0.5 1 1.5 2 0 0.5 1 1.5 2
Distance under ground, units: m Distance under ground, units: m
Wave form in zdomain, at time t=0.0054744 Wave form in zdomain, at time t=0.0068447
4 4
x 10 x 10
8
8 7
6
6
5
4 4
3
2
2
0 1
0
0 0.5 1 1.5 2 0 0.5 1 1.5 2
Distance under ground, units: m Distance under ground, units: m
Figure 24: Wave form in the soil column at various times (the buried object location is represented by
vertical dashed line).
in the small amount of extra displacement to the right of the dashed line) and having
some energy reflected. The basic one dimensional model we have used can effectively
model the presence of a rigid body in a column of soil, in particular the behavior of
partial reflecting and transmitting of energy when the object is impacted.
This numerical example has demonstrated the ability of a one dimensional model
to capture some of the salient features of wave propagation in the soil medium, in-
cluding sensitivity to soil structure and to the presence of rigid objects in the soil.
Additional features that would require a two or three dimensional model might be
modeling the presence of more than one type of body wave (e.g., shear and compres-
sional) as well as modeling surface waves. In such a higher dimensional setting we
could also take into account more complicated buried object geometries. Ultimately,
the one dimensional model still captures much of the basic dynamics of elasticity in
soil and is therefore useful in predicting outcomes to physical experiments with some
degree of fidelity.
Acknowledgments
This research was supported in part by the Air Force Office of Scientific Research
under grant number FA9550-09-1-0226. The efforts of ZRK were supported in part
by the Department of Education with a GAANN Fellowship under grant number
P200A070386. The authors are grateful to Dr. Richard Albanese for encouragement,
suggestions and constructive comments during the course of preparation of the mate-
rial in this manuscript.
References
[1] H. T. B ANKS , J. H. B ARNES , A. E BERHARDT, H. T RAN AND S. W YNNE, Modeling and
computation of propagating waves from coronary stenoses, Comput. Appl. Math., 21 (2002),
pp. 767788.
[2] H. T. B ANKS , J. B. H OOD , N. G. M EDHIN AND J. R. S AMUELS, A stick-slip/Rouse hybrid
model for viscoelasticity in polymers, Technical Report CRSC-TR06-26, NCSU, November,
2006, Nonlinear. Anal. Real., 9 (2008), pp. 21282149.
[3] H. T. B ANKS AND N. L UKE, Modelling of propagating shear waves in biotissue employing an
internal variable approach to dissipation, Commun. Comput. Phys., 3 (2008), pp. 603640.
[4] H. T. B ANKS , N. G. M EDHIN AND G. A. P INTER, Nonlinear reptation in molecular based
hysteresis models for polymers, Quart. Appl. Math., 62 (2004), pp. 767779.
[5] H. T. B ANKS , N. G. M EDHIN AND G. A. P INTER, Multiscale considerations in modeling
of nonlinear elastomers, Technical Report CRSC-TR03-42, NCSU, October, 2003, J. Comp.
Meth. Engr. Sci. Mech., 8 (2007), pp. 5362.
[6] H. T. B ANKS , N. G. M EDHIN AND G. A. P INTER, Modeling of viscoelastic shear: a non-
linear stick-slip formulation, CRSC-TR06-07, February, 2006, Dyn. Sys. Appl., 17 (2008), pp.
383406.
[7] H. T. B ANKS AND G. A. P INTER, Damping: hysteretic damping and models, CRSC-TR99-36,
NCSU, December, 1999; Encyclopedia of Vibration (S. G. Braun, D. Ewins and S. Rao,
eds.), Academic Press, London, 2001, pp. 658664.
[8] H. T. B ANKS AND G. A. P INTER, A probabilistic multiscale approach to hysteresis in shear
wave propagation in biotissue, Mult. Model. Sim., 3 (2005), pp. 395412.
[9] H. T. B ANKS , G. A. P INTER , L. K. P OTTER , M. J. G AITENS AND L. C. YANYO, Modeling
of quasistatic and dynamic load responses of filled viscoelastic materials, CRSC-TR98-48, NCSU,
December, 1998; Chapter 11 in Mathematical Modeling: Case Studies from Industry (E.
Cumberbatch and A. Fitt, eds.), Cambridge University Press, 2001, pp. 229252.
[10] H. T. B ANKS , G. A. P INTER , L. K. P OTTER , B. C. M UNOZ AND L. C. YANYO, Estimation
and control related issues in smart material structures and fluids, CRSC-TR98-02, NCSU, Jan-
uary, 1998, Optimization Techniques and Applications (L. Caccetta, et al., eds.), Curtain
Univ. Press, July, 1998, pp. 1934.
[11] H. T. B ANKS AND J. R. S AMUELS , J R ., Detection of cardiac occlusions using viscoelastic wave
propagation, CRSC-TR08-23, North Carolina State University, 2008; AAMM., 1 (2009), pp.
128.
[12] H. T. B ANKS, A brief review of some approaches to hysteresis in viscoelastic polymers, CRSC-
TR08-02, January, 2008; Nonlinear. Anal. Theor., 69 (2008), pp. 807815.
[13] J. P. B ARDET, A viscoelastic model for the dynamic behavior of saturated poroelastic soils, J.
Appl. Mech. ASME., 59 (1992), pp. 128135.
[14] R. B. B IRD , C. F. C URTISS , R. C. A RMSTRONG AND O. H ASSAGER, Dynamics of Poly-
meric Liquids, Vol. 2, Kinetic Theory, Wiley, New York, 1987.
[15] M. A. B IOT, Theory of propagation of elastic waves in a fluid saturated porous solid, J. Acust.
Soc. Am., 28 (1956), pp. 168191.
[16] R. M. C HRISTENSEN, Theory of Viscoelasticity, 2nd ed., Academic Proess, New York,
1982.
[17] N. C HOUW AND G. S CHMID, Wave Propagation, Moving Load, Vibration Reduction:
Proceedings of the International Workshop WAVE 2002, Okayama, Japan, 18-20 Septem-
ber 2002, Taylor & Francis, 2003.
[18] M. A. D EL N OBILE , S. C HILLO , A. M ENTANA AND A. B AIANO, Use of the generalized
Maxwell model for describing the stress relaxation behavior of solid-like foods, J. Food. Eng., 78
(2007), pp. 978983.
[19] M. D OI AND M. E DWARDS, The Theory of Polymer Dynamics, Oxford, New York, 1986.
[20] C. S. D RAPACA , S. S IVALOGANATHAN AND G. T ENTI, Nonlinear constitutive laws in vis-
coelasticity, Math. Mech. Solids., 12 (2007), pp. 475501.
[21] J. D. F ERRY, E. R. F ITZGERALD , L. D. G RANDINE AND M. L. W ILLIAMS, Temperature
dependence of dynamic properties of elastomers: relaxation distributions, Ind. Engr. Chem., 44
(1952), pp. 703706.
[22] W. N. F INDLEY AND J. S. Y. L AI, A modified superposition principle applied to creep of nonlin-
ear viscoelastic materials under abrupt changes in state of combined stress, Trans. Soc. Rheol.,
11 (1967), pp. 361380.
[23] W. N. F INDLEY, J. S. L AI AND K. O NARAN, Creep and Relaxation of Nonlinear Vis-
coelastic Materials, Dover Publications, New York, 1989.
[24] Y. C. F UNG, Foundations of Solid Mechanics, Prentice-Hall, Englewood Cliffs, NJ, 1965.
[25] Y. C. F UNG, Biomechanics: Mechanical Properties of Living Tissue, Springer-Verlag,
Berlin, 1993.
[26] Y. C. F UNG, A First Course in Continuum Mechanics, Prentice Hall, New Jersey, 1994.
[27] A. E. G REEN AND R. S. R IVLIN, The mechanics of non-linear materials with memory, Arch.
Ration. Mech. An., 1 (1957), pp. 121.

[28] M. G URTIN AND E. S TERNBERG, On the linear theory of viscoelasticity, Arch. Ration. Mech.
An., 11 (1965), pp. 291365.
[29] D. T ER H AAR, A phenomenological theory of viscoelastic behavior, Phys., 16 (1950), pp. 839
850.
[30] Y. M. H ADDAD, Viscoelasticity of Engineering Materials, Chapman & Hall, Dordrecht,
Netherlands, 1995.
[31] B. O. H ARDIN, The nature of damping in sands, J. Soil Mech. Found. Div., 91 (1965), pp.
6397.
[32] A. R. J OHNSON AND R. G. S TACER, Rubber viscoelasticity using the physically constrained
systems stretches as internal variables, Rubber. Chem. Technol., 66 (1993), pp. 567577.
[33] R. S. L AKES, Viscoelastic Solids, CRC Press, Boca Raton, FL, 1998.
[34] H. L EADERMAN, Elastic and Creep Properties of Filamentous Materials, Textile Founda-
tion, Washington, D.C., 1943.
[35] P. L E TALLEC, Numerical Analysis of Viscoelastic Problems, Masson/Springer-Verlag,
Paris/Berlin, 1990.
[36] P. L E TALLEC , C. R AHIER AND A. K AISS, Three-dimensional incompressible viscoelasticity
in large strains: formulation and numerical approximation, Comput. Meth. Appl. Mech. Eng.,
109 (1993), pp. 233258.
[37] F. J. L OCKETT, Nonlinear Viscoelastic Solids, Academic, New York, 1972.
[38] N. L UKE, Modeling Shear Wave Propagation in Biotissue: An Internal Variable Ap-
proach to Dissipation, Ph.D. thesis, North Carolina State University, Raleigh, NC,
http://www.ncsu.edu/grad/etd/online.html, 2006.
[39] J. E. M ARSDEN AND T. J. R. H UGHES, Mathematical Foundations of Elasticity, Prentice-
Hall, Englewood Cliffs, NJ, 1983.
[40] P. M ICHAELS, Relating damping to soil permeability, Int. J. Geomech., 6 (2006), pp. 158165.
[41] P. M ICHAELS, Water, inertial damping, and complex shear modulus, Geotech. SP., 2008.
[42] K. N. M ORMAN . J R ., Rubber viscoelasticity-a review of current understanding, Proceedings
of the Second Symposium on Analysis and Design of Rubber Parts, Jan. 1415, 1985.
[43] J. G. O AKLEY, A. J. G IACOMIN AND J. A. Y OSICK, Molecular origins of nonlinear viscoelas-
ticity: fundamental review, Mikrochimica. Acta., 130 (1998), pp. 128.
[44] R. W. O GDEN, Non-Linear Elastic Deformations, Dover Publications, 1984.
[45] A. P. P ENNERU , K. J AYARAMAN AND D. B HATTACHARYYA, Viscoelastic behaviour of solid
wood under compressive loading, Holzforschung., 60 (2006), pp. 294298.
[46] A. C., P IPKIN AND T. G. R OGERS, A non-linear integral representation for viscoelastic be-
haviour, J. Mech. Phys. Solids., 16 (1968), pp. 5972.
[47] F. E. R ICHART AND J. R. W OODS, Vibrations of Soils and Foundations, Prentice Hall, NJ,
1970.
[48] R. S. R IVLIN, Large elastic deformations of isotropic materials, I, II, III, Phil. Trans. Roy. Soc.
A., 240 (1948), pp. 459525.
[49] P. E. R OUSE . J R ., A theory of the linear viscoelastic properties of dilute solutions of coiling
polymers, J. Chem. Phys., 21 (1953), pp. 12721280.
[50] J. R. S AMUELS, Inverse Problems and Post Analysis Techniques for a Stenosis-Driven
Acoustic Wave Propagation Model, Ph.D. thesis, North Carolina State University,
Raleigh, NC, http://www.ncsu.edu/grad/etd/online.html, 2008.
[51] R. A. S CHAPERY, On the characterization of nonlinear viscoelastic solids, Polymer. Eng. Sci.,
9 (1969), pp. 295310.
[52] R. A. S CHAPERY, Nonlinear viscoelastic and viscoplastic constitutive equations based on ther-
modynamics, Mech. Time-Depend. Mat., 1 (1997), pp. 209240.

[53] R. A. S CHAPERY, Nonlinear viscoelastic solids, Int. J. Solids. Struct., 37 (2000), pp. 359366.
[54] C. T. S CHRODER , W. R. S COTT, J R ., AND G. D. L ARSON, Elastic waves interacting with
buried land mines: a study using the FDTD method, IEEE Trans. Geosci. Remote. Sens., 40
(2002), pp. 14051415.
[55] F. S CHWARZL AND A. J. S TAVERMAN, Higher approximation methods for the relaxation spec-
trum from static and dynamic measurements of viscoelastic materials, Appl. Sci. Res., A4 (1953),
pp. 127141.
[56] W. R. S COTT, G. D. L ARSON AND J. S. M ARTIN, Simultaneous use of elastic and electromag-
netic waves for the detection of buried land mines, Proc. SPIE., 4038 (2000), pp. 112.
[57] T. H. S HELLHAMMER , T. R. R UMSEY AND J. M. K ROCHTA, Viscoelastic properties of edible
lipids, J. Food. Eng., 33 (1997), pp. 305320.
[58] J. S MART AND J. G. W ILLIAMS, A comparison of single integral non-linear viscoelasticity
theories, J. Mech. Phys. Solids., 20 (1972), pp. 313324.
[59] L. R. G. T RELOAR, The Physics of Rubber Elasticity, 3rd edition, Oxford University Press,
1975.
[60] C. T RUESDELL, Rational Thermodynamics, Springer, Berlin, 1984.
[61] A. W INEMAN, Nonlinear viscoelastic solids-a review, Math. Mech. Solids., 14 (2009), pp.
300366.

A Brief Review of Elasticity and Viscoelasticity For Solids: Eview Rticle

Uploaded by

Copyright:

Available Formats

A Brief Review of Elasticity and Viscoelasticity For Solids: Eview Rticle

Uploaded by

Document Information

Original Description:

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

A Brief Review of Elasticity and Viscoelasticity For Solids: Eview Rticle

Uploaded by

Copyright:

Available Formats

Advances in Applied Mathematics and Mechanics DOI: 10.4208/aamm.

Abstract. There are a number of interesting applications where modeling elastic

2 Preliminary notions and balance laws

1. A rigid-body displacement: in this case the relative displacement between par-

2.1.2 Displacement and strain

The (Lagrangian) displacement of the particle relative to X is given by

U(X) = x(X) X. (2.3)

If we assume the transformation has a unique inverse, then we have

Figure 1: Deformation of a body.

The Lagrangian strain tensor

The Lagrangian (finite) strain tensor E is defined by

where I is the identity matrix, and U = (U1 , U2 , U3 ) T . Thus, because

the relationship between Lagrangian strain (2.6) and displacement is given by

where Eij is the (i, j) component of strain tensor E.

The Eulerian strain tensor

|dx|2 |dX|2 =(dx)T dx (dx)T ( A1 )T A1 dx

where eij is the (i, j) component of strain tensor e.

where i , i = 1, 2, 3 are the eigenvalues of A, and also known as principal stretches

Infinitesimal strain theory

Definition 2.4. The stress vector (traction) is defined by

Figure 2: Notations of stress components.

Definition 2.5. The Cauchy stress tensor is defined by

Figure 3: Stress vector acting on a plane with normal n.

Here ( (n) (n) (n) ) T

2.2 Equations of motion of a continuum

2.2.1 The material derivative of a volume integral

where on the boundary t of t , v = v(t) is the velocity

2.2.3 The Reynolds transport theorem

where Dvz /Dt is the total derivative of vz , and is given by

2.2.4 The Eulerian equations of motion of a continuum

Then by Gauss Theorem and the above equation we find that

3 Constitutive relationships: stress and strain

to displacements. To complete our derivations of the equations of motion, we must

Dijkl = ij kl + (ik jl + il jk ). (3.1)

3.1 Elastic materials

3.1.1 Linear elasticity

3.1.2 Nonlinear elasticity

Nonlinear stress-strain relations

Strain energy function for isotropic elastic materials

Special cases for (3.9) include several materials:

1. A Neo-Hookean material for which W ( I1 , I2 ) = c10 ( I1 3),

W ( I1 , I2 ) = c10 ( I1 3) + c01 ( I2 3).

3.2 Viscoelastic materials

3.2.1 Properties of viscoelastic materials

lim G (t) = G > 0,

Figure 4: Stress and strain histories in the stress relaxation test.

Figure 5: Stress and strain histories in the creep test.

while for viscoelastic fluids the stress vanishes to zero, i.e.,

lim J (t) = J > 0,

(t) = 0 sin(t + ), (3.11)

(t) = 0 cos() sin(t) + 0 sin() cos(t). (3.12)

F (t) = a (t) = a0 cos() sin(t) + a0 sin() cos(t).

The above equation can be rewritten as

The characteristic parameter G is referred to as the complex dynamic modulus. We

The relations may be inverted to obtain

3.2.3 Linear viscoelastic models: constitutive relationships

Figure 8: Schematic representation of the Maxwell model.

The Maxwell model