You are on page 1of 22

Introduction to Multiphysics

Lecture Notes

Jagan M. Padbidri

School of Engineering Sciences


Mahindra École Centrale

Fall 2017
Introduction
The previous section of the course was devoted to understanding the principles of continuum
mechanics. In order to pursue these studies, we represented objects as continuous bodies and the
relevant quantities of interest as fields described over these continuous objects. These fields usually
represent physical quantities of interest, for example, the displacement field of an object is useful to
understand the deformation undergone by an object. These fields could be scalar fields as well,
temperature, for example. Using such a continuous description of objects and fields, we developed
expressions that describe the fundamental laws of physics such as conservation of mass, momenta and
energy. We recall that these laws are also termed as the Principles of mechanics and these laws
govern the overall behavior and evolution of a system. In this section of the course, we will address
the application of these governing laws to describe specific physical phenomena of interest.

Our primary objective in this part of the course will be to develop the methodology of
representing different physical processes in the continuum framework. We have previously addressed
that the behavior of the system, as a whole, is described by the governing equations for the relevant
physical process. Further, the quantities that describe the state of the object as it goes through a
physical process evolve according to Constitutive relations. For example, if we consider the simple
case of an object deforming according to some external forces and constraints, the state of
deformation at any physical point would be described by the strain tensor at that point. The response
of the material due to the applied forces and constraints is described by the stress tensor at any given
material point. These two quantities describe the deformation state of the object as it undergoes the
deformation process. The framework that relates these two tensorial quantities is referred to as the
Constitutive relation for deformation. To complete the description of a physical process, one also
needs to define the quantities that drive a process and/ or restrict the physical process. Continuing
with the example of an object deforming under a set of external forces and constraints, it is
straightforward to understand that there would not be any deformation in the absence of external
forces. Thus, the applied external forces drive the deformation process to which the body is subject. In
addition to the presence of external forces, one would also require a set of constraints which restrict
the rigid body motion of the object. For example, if no physical point on the object is fixed/
constrained to move, the object would simply exhibit rigid body acceleration for an infinite amount of
time. If we were to constrain specific material points on a body, then we would be subjecting the body
to a deformation process. Thus, to complete the description of a physical phenomenon, we would
need to specify the relevant Boundary Conditions that drive the physical process. Thus, to describe the
physics governing the evolution of a system, one would need to formulate the relevant governing
equations subject to the principles of mechanics, specify the constitutive relations and boundary
conditions.

Along with developing the framework to describe a single physical process, we also focus on
processes in which more than one physical mechanism affects the behavior of the system. The
problems/ phenomena in which more than one physical mechanism is involved are referred to
Coupled phenomena. As a simple example, we consider the flow of electricity through a conducting
wire. It is known that the flow of electricity generates heat in the conductor – the working principle of
an incandescent lamp. Also, the temperature of a conductor affects the flow of electricity in a
conducting material since the electrical resistance of a conductor increases with temperature. The
increase of electrical resistance affects the electrical current flowing through the conductor and
subsequently the amount of heat generated. Thus, we see that the two physical different phenomena of
electrical conduction and heat generation affect each other, i.e., each physical process is coupled to
the other. Such types of phenomena are known as Strongly Coupled phenomena. On the other hand,
let us consider the phenomenon of the deformation undergone by an object subject to thermal loads;
say the component of an IC engine. We do know that a substantial amount of heat is generated in such
a system and as the heat is transferred to the component, i.e., as the component gets hotter, it tends to
expand and generates stresses and strains. Thus, the thermal phenomenon is influencing the physics of
deformation. However, the fact that the object/ component is subject to deformation does not in any
way affect the generation or subsequent dissipation of heat, i.e., one physical phenomenon affects the
other, but the dependence is not mutual. These kinds of processes are referred to as Weakly Coupled
phenomena.

In our studies here, we will describe both strongly coupled and weakly coupled processes such as
Thermo-Elastic coupling, Thermal-Electrical coupling and Electrical-Elastic coupling. We emphasize
here that the primary objective here is to develop the ability to describe physical processes by
formulating the relevant governing differential equations, constitutive relations and specifying the
appropriate boundary conditions. The details of the methods to solve these problems are beyond the
scope of this course.
Chapter 1

Thermo-Elastic Phenomena

In this chapter, our objective is to address the representation of phenomena that capture the
interplay of temperature related (thermal) processes and physical deformation of objects (elastic
phenomena). However, we first present the relevant formulations for the linear elastic deformation of
objects and later consider the combination of thermal phenomena and elastic deformation.

Formulation for Linear Elasticity


Consider the deformation of an object under the influence of external forces and constraints, as
shown in Fig. 1.1. The domain of the object is defined as  with the boundary defined as  . Points
A and B on the surface of the body are rigidly constrained and cannot move and we apply some force,
in the form of traction, t over part of the boundary t . The body deforms in response to the
externally imposed forces producing some displacement, and consequently strain and generates some
stresses internally, to balance the effect of the external forces. Our objective, as mentioned earlier, is
to describe the deformation process by using the relevant governing equations, constitutive relations
between the variables and the boundary conditions that drive the deformation process.

Figure 1.1. Schematic of a body subject to displacement constraints and external tractions.

In the previous chapter, we have seen that using the principle of balance of linear momentum, the
object is required to satisfy the equation
D
    x, t   b  x, t     x, t  v  x, t 
Dt
We can simplify the representation of the above equation by dropping the explicit mention of the
position vector, x and time t for the terms. Also, since the velocity of a point is the rate of change of
the displacement, the above equation takes the form

d 2u
   b  
dt 2
We impose certain requirements on the nature of the deformation by specifying that we consider a
method of deformation which is independent of inertial effects. Physically this translates to
deformation which is very slow and thus, the acceleration (inertial) term in the above equation can be
ignored to yield
   b  0 (1.1)

Further, since the object is under equilibrium with the external forces, the Eqn. (1.1) has to be
satisfied at each point in the body and thus, forms the governing differential equation for equilibrium
static deformation of an object.

From the above equation, it may seem that the variable that one should solve for is the stress field
over the object. However, the primary variable that we solve for is the displacement field in the body.
Once the displacement field has been obtained, the strain and stress fields can be obtained from it. In
elementary introduction to mechanics, the strain and stress in an object are related using the Elastic
modulus as   E where E is the Young’s elastic modulus. Another quantity that was introduced in
elementary mechanics is the Poisson’s ratio represented by  . However, this relation is applicable
only to one-dimensional stress-strain states which may be justified to be scalars. We have proved in
our study of continuum mechanics that the stress and strain states at a point in an object are not scalar
quantities, but are second order tensorial quantities that have multiple components and the simplistic
  E would not be extendable to all components of stress and strain. The relation between the
components of strain and stress tensors is given as

E E
 ij 
E
1  
 ij 
1  1  2 
 kk  ij or  
E

1   1  1  2 
 
tr  (1.2)

The fourth order stiffness tensor


The above equation represents the relation between the two quantities, strain and stress, which
characterize the state of deformation at a physical point. Thus, the above equation forms the linear
elastic constitutive relation for deformation of objects. However, the representation in the above form
is mathematically not elegant. Indeed, the above forms of the constitutive relations result from the
components of the Stiffness tensor, which itself is derived from thermodynamic principles. The
physical motivation for a thermodynamic approach stems from the observation that the set of external
forces acting on the body are essentially doing work on the system to increase the internal energy of
the system. We know from the balance of mechanical energy that the rate of work done by the
external forces acting on an object is given by

dWext  Dv 
   Dt  v  D :   dV
dt  t
 

Dv
In the above equation, the first term   v represents the kinetic energy of the infinitismal volume
Dt
and the second term, D :  represents the stress power which serves as a representative of the rate at
which the internal energy of the object changes due to deformation. In the context of describing
relations for thermo-elastic phenomena, the total internal energy of an object can be assumed to be
composed of internal energies accounting for both mechanical deformation and thermal phenomena.
However, here we isolate the internal energy due to mechanical deformation – the strain energy
density, the rate of change of which is given by the stress power, i.e., if we represent the strain energy
density using a function  , then   D :  , where D is the rate of deformation tensor. We have
restricted our previous discussions of describing the deformation of an object to being limited to
infinitismal strain. For such cases, the above relation can be expressed as

d
  D :   :   :  (1.3)
dt
We postulate that the internal energy is a function of the deformation state of the object, i.e., it is
exclusively function of the strain state of the object. Further, to define the constitutive relations for
linear elastic materials, we define the energy function as a quadratic function of the strain state, i.e.,

  1
     A  B:    : C : 
2
(1.4)

We note here that the function Ψ represents energy which is a scalar quantity. Thus, each term of Eqn.
(1.4) has to be a scalar quantity. We can infer from this requirement that the quantity B is a second
order tensor and C is a fourth order tensor. Further, both B and C represent properties of the
material, rather than the state of the deformation of the object. From Eqn. (1.4), knowing that Ψ is a
function of strain, we can write


d      : d    :  (1.5)
dt  dt 

Comparing the above equation with Eqn. (1.3), we obtain

 
   ij  (1.6)
  ij

Using this relation between the strain energy function Ψ and the stress tensor, from Eqn. (1.4), we can
write

  1 
  A  B:   :C: 
  2 
  1 
  mn   A  Bij  ij   ij Cijkl  kl 
 mn  2 
 ij 1  ij 1 
 mn  Bij  Cijkl  kl   ij Cijkl kl
 mn 2  mn 2  mn

 mn  Bij  im jn    
1 1
 im jn Cijkl  kl   ij Cijkl  km ln 
2 2
1 1 1 1
 mn  Bmn  Cmnkl  kl  Cijmn ij  Bmn  Cmnkl  kl  Cklmn kl
2 2 2 2
If the fourth order tensor C satisfies the condition that Cmnkl  Cklmn , then the above relation
simplifies to

 mn  Bmn  Cmnkl  kl (1.7)


The above equation relates the components of strain to components of the resultant stress. It is
noteworthy that the above relation allows the possibility of the stress components being non-zero even
if the magnitude of the strain components   kl  is equal to 0. Physically, this means that the strain and
stress are perhaps being measured using different points of reference during the process of
deformation. To avoid such ambiguities, we impose the condition that at a reference state when the
strain components are 0, the magnitudes of the stress tensor components are also 0, i.e., Bmn  0 . This
simplifies Eqn. (1.7) to

 mn  Cmnkl  kl or   C :  (1.8)

The above equation represents a mathematically much more elegant form than the stress-strain
relation shown in Eqn. (1.2). Further, we have derived this relation from thermodynamic arguments
maintaining consistency with the balance of energy. The quantity C is called the Stiffness tensor,
which is a material property and is a fourth order tensor with 81 components.

While the stress-strain relation obtained in Eqn. (1.8) is mathematically elegant, the representation
of the stiffness tensor C could be rather cumbersome, owing to the possibility of 81 different
components. However, there are certain symmetries that the stiffness tensor would need to satisfy
since it relates the symmetric stress and strain tensors, i.e, since the stress tensor is symmetric,
 ij   ji  Cijkl  kl  C jikl  kl  Cijkl  C jikl . Similarly, due to symmetry of the strain tensor, we
obtain Cijkl  Cijlk . To satisfy these constraints, the number of independent components needed to
describe the stiffness tensor, C reduces from 81 to 36. These symmetries induced in the stiffness
tensor due to the symmetry of the stress and strain tensors are referred to as minor symmetries. In
addition to the symmetry due to the stress and strain tensors, the stiffness tensor has additional
symmetry due to being derived from a scalar energy potential functional, i.e., Cijkl  Cklij . This is
referred to as a major symmetry of the stiffness tensor and it further reduces the number of
independent components of the stiffness tensor to 18. The expansion of Eqn. (1.8) simplifies to a great
extent due to the above mentioned symmetry of the stiffness tensor. Ideally, a fourth order tensor
would require a 4D data structure to represent its components. However, taking advantage of the
lesser number of independent components of C , this can be represented using a matrix structure.
Further, utilizing the symmetry of the stress and strain tensors, they can be represented using a 1D
data structure. The matrix representation of Eqn. (1.8) can then be given as

 11  C1111 C1122 C1133 C1123 C1113 C1112   11 


   C2222 C2233 C2223 C2213 C2212    22 
 22  
 33   C3333 C3323 C3313 C3312    33 
    (1.9)
 23   C2323 C2313 C2312   2 23 
 13   Symmetric C1313 C1312   213 
    
 12   C1212   212 

To simplify this representation further, we introduce a notation that contracts a pair of indices of the
stiffness tensor using the following convention 11  1 , 22  2 , 33  3 , 23  4 , 13  5 and
12  6 . This notation, i.e., the contraction of indices is called as the Voigt notation which simplifies
the representation of the components of the stiffness tensor in Eqn. (1.9) to
 11  C11 C12 C13 C14 C15 C16   11 
   C22 C23 C24 C25 C26    22 
 22  
 33   C33 C34 C35 C36    33 
    (1.10)
 23   C44 C45 C46   2 23 
 13   Symmetric C55 C56   213 
    
 12   C66   212 

Using the Voigt notation to the components of the stress and strain tensors as well, we arrive at

 1  C11 C12 C13 C14 C15 C16   1 


   C22 C23 C24 C25 C26    2 
 2 
 3   C33 C34 C35 C36    3 
    (1.11)
 4   C44 C45 C46   2 4 
 5   Symmetric C55 C56   2 5 
    
 6   C66   2 6 

Note: Contracting the indices of the above quantities does not alter their fundamental nature, i.e.,
representing the stresses and strains using a single index in Eqn. (1.11) does not imply that they are
vectors. These quantities fundamentally represent second order tensors and the components transform
as the components of a second order would. Similarly, using two indices to represent the components
of the stiffness tensor does not render it as a second order tensor. The above notation has been used
purely for convenience of representation.

In addition to the number of independent components of the stiffness tensor decreasing due to the
symmetry of stress and strain, a large number of components are rendered redundant due to the
symmetry of material structure. Recall from ME 203 that the structure of all solids is described by the
pattern in which atoms are arranged in a lattice. Thus, at the atomic level, the properties of a material
are very much dependent on the sequence by which the atoms are stacked. This is reflected as
anisotropy of the material properties at the continuum length scale with properties in different
directions varying due to the discrete nature of the atomic packing. This anisotropy in material
properties extends until a grain in the material with the orientation of the grain determining the
direction of anisotropy. However, as we consider a larger length scale, i.e., a physical component
containing millions of such grains, the effect of the anisotropy is smoothened out since grains of all
possible orientation are encountered and the anisotropy due to atomic packing seems inconsequential
and the material displays the same physical properties in all directions. This property of the material is
called as isotropy. For an isotropic material, the elastic stiffness can be completely described by two
independent constants, much lower than the aforementioned 18. Thus, two independent quantities
form the components of the stiffness tensor for isotropic materials. Previously, we have seen that in
the elementary treatment of mechanics, these two constants are given by the elastic Young’s modulus
(E) and the Poisson’s ratio (ν). These two quantities can be represented in the stiffness tensor C to
describe the elastic properties of isotropic materials. The general form of the components of the
stiffness tensor for isotropic materials is given as
 11   C11 C12 C12 0 0 0   11 
  C 0    22 
 22   12 C11 C12 0 0
 33  C12 C12 C11 0 0 0    33 
    (1.12)
 23   0 0 0 C44 0 0   2 23 
 13   0 0 0 0 C44 0   213 
    
 12   0 0 0 0 0 C44   212 

where only C11 and C12 are independent components of the stiffness tensor and C44 = (C11 - C12)/2 is
not an independent quantity. The components of the stiffness tensor, and the quantities E and ν are
related as

1  
C11  E
1   1  2 
(1.13)

C12  E
1   1  2 
Thus, having detailed the constitutive relations for linear elasticity, we can write the formulation for
the problem shown in Fig. 1.1 as
   b  0 the governing equation.
1
  C:  u  u  , and
the constitutive relation, where   (1.14)
2
u A  u B  0, nˆ    t on t are the boundary conditions.

Formulation for Linear Thermo-Elasticity


In the previous section, we have formulated the relevant equations, relations and constraints to
describe the mechanical deformation of a linear elastic material. Here, we address the formulation of
the boundary value problem for a thermo-elastic process. Thermo-elastic/ thermo-mechanical
processes include both thermal phenomena and mechanical deformation. We have previously argued
that thermo-elasticity is an example of a weakly coupled problem since the thermal process affects the
mechanical deformation via generating thermal strain (expansion/ contraction), but the mechanical
deformation does not affect the thermal process. For such processes, we have already described the
governing equations for the mechanical deformation using the balance of linear momentum and the
governing equation for the thermal process using the balance of thermal energy and correspondingly,
the heat equation. We will need to develop relations between the variables that are used to quantify
this process, i.e., we will need to develop the relevant constitutive relations.

We aim to describe the process undergone by an object that is subject to both thermal and
mechanical loads, as shown in Fig. 1.2. The domain of the object is defined as  with the boundary
defined as  . For the mechanical deformation, points A and B on the surface of the body are rigidly
constrained and cannot move, and traction, t is applied over part of the boundary, t . In addition to
the mechanical constraints, we also subject the body to thermal loads by prescribing that part of the
boundary, T is prescribed a temperature T0 and a heat flux of magnitude q0 is prescribed on part of
the boundary defined by  q .
T t
T0

B t
Figure 1.2. Schematic of a body
ê3
subject to a thermo-elastic process.
A
ê1 ê2
q0

We realize that in such a case, one would need to solve for solution variables that describe both
the mechanical and thermal processes, i.e., one would need to solve for the temperature field and the
displacement field. The displacement field will need to satisfy the governing equation for the
 
mechanical deformation     b  0 and the temperature field (T) will need to satisfy the balance

 T 
of thermal energy   c p      T   Q  where cp is the specific heat capacity of the material,
 t 
 is the second order thermal conductivity tensor and Q is the thermal energy generated per unit
volume in the object. Thus, forming the differential equations that govern both the mechanical and
thermal phenomena is fairly straightforward. Further, describing the boundary conditions for both
these phenomena is also relatively simple. We will, however, need to pay attention to forming the
constitutive relations for the thermo-mechanical process. For an isolated mechanical deformation
process, we have derived the relevant constitutive relations as   C :  , given by Eqn. (1.8) and for
an isolated thermal process, we know that the constitutive relations are given by Fourier’s law as
q     T  where q is the heat flux vector and T is the temperature field. However, in the context
of a multiphysics problem such as thermo-elasticity, these relations will not all be directly extendable.

For example, consider the deformation of an object subject to both mechanical and thermal
loading. The mechanical loading imposed as tractions introduces a displacement field in the object
which results in strain and consequently, stress. However, this is not the only mechanism by which
strain and stresses are generated. When a body is subject to thermal loading resulting in variations of
temperature, that body would tend to undergo volumetric expansion/ contraction. If the imposed
displacement constraints prevent a free expansion, then stresses are generated in the object due to
application of thermal loads. Thus, the stress state of the object is dependent not only on the strains
generated due to the mechanical loading, but also due to the temperature variations corresponding to
thermal loading. We will have to then revise the definition of the function  which has been used to
derive the stiffness tensor and the constitutive relations. For thermo-mechanical processes, the
function  is defined to be dependent on both the strain and the temperature (T), i.e.,     ,T .  
While not within the scope of the material presented here, we can use the second law of
thermodynamics to prove that this definition of  can satisfy Eqn. (1.6), i.e.,

 
   ij  (1.15)
  ij
To obtain the relation between stress, strain and temperature, we define  to be a function of both
strain and temperature. We can express the function  as a second order Taylor series about a
reference temperature Tref and a reference strain  ref , i.e.,

 
  
  , T    ref , Tref  


:    ref  
T  ref ,Tref

T  Tref 
 ref ,Tref

1 2
 T
2
T  Tref    
2
  T  Tref :    ref (1.16)
2 T 2 
ref ,Tref  ref ,Tref

2

1
2
   ref : 2

 
:    ref 
 ref ,Tref

From the definition of stress given in Eqn. (1.15), we get

  2  2





T
 T  Tref  
 2

:    ref  (1.17)
 ref ,Tref  ref ,Tref  ref ,Tref

Similar to the approach followed before, we define deformation from a state where the strain is 0, i.e.,
 ref  0 and require the initial stress   0 . Then,

2  2

 2
: 
T
T  Tref  (1.18)
Tref Tref

We know that  2   2 is a fourth order tensor, which we have represented as the stiffness tensor C .

We represent  2   T as a second order tensor β . Then the stress is given by

  C :   β T  Tref    ij  Cijkl  kl  ij T  Tref   (1.19)

Writing the tensor β in terms of the stiffness tensor C as β  C : α  ij  Cijkl kl , we can express
the stress as

  C :   α T  Tref    ij  Cijkl  kl   kl T  Tref  (1.20)

We note here that the stress – strain relation in Eqn. (1.20) is similar in structure to the linear
elastic relation of Eqn. (1.8) , but the above equation also contains the effect of temperature on stress.
This additional term physically represents the strains caused due to thermal expansion, i.e.,
α represents the second order thermal expansion tensor and T  Tref is the temperature change at a  
 
point in the object. Thus, α T  Tref gives the strain that results from thermal expansion. In
elementary physics, one would have been introduced to the coefficients of thermal expansion as scalar
quantities, but we find here that it is in fact a second order tensor. It is important to note here that  is
derived from the displacement field and does not distinguish if those displacements have resulted
from thermal expansion or mechanical deformation – it is merely the strain that is observed in the
object as a result of the thermo-mechanical process.
To elucidate this point further, let us consider a couple of examples. Let us assume that a body is
subject to purely thermal loading with mechanical constraints. Say for example that the temperature of
an object is increased without any physical constraints on the body, i.e., the body is free to expand. In
such a case, the strain that the body is observed to undergo is purely due to the thermal expansion, i.e.,
  α T  Tref  . Then, the stress in the body is evaluated to be equal to 0, from Eqn. (1.20), which is
physically accurate. On the contrary, let us consider a case where the temperature of a body is
increased, which should result in thermal expansion, but the body is constrained such that the
displacement at every point is 0, i.e., we are forcibly preventing the body from expanding. In this
 
case,   0 and from Eqn. (1.20), the stress is given by   C : α T  Tref , i.e., compressive stresses
have been generated in the body as a result of preventing the expansion, which also makes physical
sense. Thus, we verify that the constitutive relations do capture the effect of temperature on the
stresses and strains generated due to thermal and mechanical processes.

We have previously mentioned that the governing equations for the thermal and mechanical
processes stem from the balance of energy and the balance of linear momentum respectively. We can
now formulate the boundary value problem for the thermo-mechanical process shown in Fig. 1.2 as

   b  0 

T 
 
Governing equations.
cp    q  Q
t 

  C :   α T  Tref    1
 Constitutive relations, where    u  u  , and
q     T   2

u A  u B  0, nˆ    t on t 
 Boundary conditions.
T  T0 on T ,     T    nˆ  q0 on  q 

Note: We emphasize here that the constitutive equations derived in this chapter use the existence of a
thermodynamic function,  and the ability to derive the stress as a derivative of this function with
respect to strain. It is not necessary that all materials conform to these assumptions. However, there
are materials that do follow these relations. The class of materials for which  represents the
Helmholtz free energy and consequently, stress can be derived from this function are called as
Hyperelastic materials and are the easiest materials to deal with when describing the thermodynamics
of deformation.
Chapter 2

Thermo-Electric Phenomena
The previous chapter dealt with a combination of thermal process and mechanical deformation. In
this chapter, we address the combination of thermal and electrical process, more specifically, we will
describe the Joule heating effect. Joule heating is the process in which the electric current passing
through a conductor generates heat in the conductor. The heat generated during the flow of electric
current raises the temperature of the conductor and consequently increases the electrical resistance of
the conductor. This in turn affects the flow of electric current in the conductor. Thus, unlike the
previous example of weakly coupled phenomenon such as thermo-elasticity, the coupling between the
thermal and electric phenomena is two-way, i.e., the present example is a strongly coupled
phenomenon. Over the past few decades, modelling such phenomenon have found increasing
applications in areas such as reliability assessment of integrated circuit design.

To describe the formulation of the thermo-electric phenomenon, we will need to address the
governing equations that describe both the electric flow and the balance of thermal energy in the
system. A schematic of the thermo-electric phenomenon is shown in Fig. 2.1. The driving force for
the flow of electricity through an object is a difference in the applied electric potential. Here, we
represent the scalar electric potential using  , which is prescribed specific values on the boundary of
the object as shown below. In addition, we also have a temperature and heat flux conditions on the
boundary of the object, which drive the thermal phenomenon.

T 0

T0
0

0
Figure 2.1. Schematic of a
ê3 coupled thermo-electric problem.

ê1 ê2
q0

As we have seen in thermo-elasticity, the thermal phenomenon can be described by the balance of
 T 
energy equation, i.e.,   c p      T   Q  . In the previous chapter, the heat generation term,
 t 
 Q  did not always take a fixed value, either zero or non-zero. The value of Q was dependent on the
kind of problem that was being modeled using the principles of thermo-elasticity. In this case,
however, the generation of heat due to the flow of electricity is a fundamental characteristic of the
problem. Thus, we realize that while describing the Joule heating problem, Q has to be non-zero and
we derive the relevant representation at a later stage. While the thermal process can be described by
the balance of energy, we will also need to describe the process of electric flow. This is done using the
principle of conservation of charge and the consequent continuity equation applicable to electric
charges. This equation is given as
 free
   J (2.1)
t
Here, ρfree is the density of the free charge in an infinitismal volume and J is the vector describing the
electric current density flowing in that infinitismal volume. If we consider a steady current implying
that charge does not accumulate at any point in space with respect to time, i.e., the rate of
accumulation of free charge is 0, then the above equation simplifies to
 J  0 (2.2)
From elementary electromagnetic theory, we know that the current density vector in a volume is
related to the electric field applied across that volume and the electrical conductivity of the object,
given by
J  E (2.3)
Here, E is the electric field applied across the infinitismal volume and  is the second order electric
conductivity tensor. We again observe that electric conductivity, which is a material property and has
often been assumed to be a scalar quantity, is in fact a second order tensor which is capable of
describing anisotropy in electric conductivity. We again draw on elementary electromagnetic theory
to relate the electric field to an applied potential across the object as
E   (2.4)
where  is the applied scalar electric potential. Using the above equations, we can describe the
electric flow in the object as being governed by
 
    E  0, where E   (2.5)
This equation, in conjunction with the balance of thermal energy forms the set of governing equations
for the thermo-electric process, i.e., the thermo-electric problem can now be described as
 
  E  0 

T  Governing equations.
cp     q   Q
t 

E   
 Constitutive relations
q     T  

  0 on  0 ,   0 on  
 Boundary conditions.
T  T0 on T ,     T    nˆ  q0 on  q 

We will still need to describe the rate at which heat is generated due to the flow of electricity. We
know that this is given by the product of the current and the applied potential difference, per
elementary electromagnetics. In the current context, the rate of heat generation is given as the product
 
of the electric field and the current density, i.e., Q  E  J         . The problem description
for the thermo-electric phenomenon can now be given by
 
  E  0 

T  Governing equations.
cp     q            
t 

E   
 Constitutive relations (2.6)
q     T  

  0 on  0 ,   0 on  
 Boundary conditions.
T  T0 on T ,     T    nˆ  q0 on  q 
In the above formulation, we have considered the effect of the electric flow on the generation of heat,
but we have not explicitly accounted for the effect of the thermal phenomenon on the electric flow,
i.e., in the form given in Eqn. (2.6) we are describing the problem as a weakly coupled phenomenon.
In reality, the increase in temperature of the affects the electric conductivity of the material, i.e.,
   T  , then the above equation can be written as

 
   T   E  0 

T  Governing equations.
cp     q          T      
t 

E   
 Constitutive relations (2.7)
q     T  

  0 on  0 ,   0 on  
 Boundary conditions.
T  T0 on T ,     T    nˆ  q0 on  q 
which describes the fully coupled version of the thermo-electric phenomenon. The electrical
conductivity is a second order tensor which is represented as a 3X3 matrix in 3D space. The diagonal
terms of this matrix representation predict the current flow along the direction of the applied potential
gradient and the off-diagonal terms capture the anisotropic effects. For most engineering materials, it
can be safely assumed that the electrical conductivity is purely diagonal and can be represented by a
scalar quantity,  T  . Following this simplification, the heat generation term can be described as

Q   T   .
2
Chapter 3

Electro-Elastic Coupling: Piezoelectric


Phenomena

We now address the phenomenon in which mechanical deformation is coupled with electric
phenomenon. There are specific kinds of materials which exhibit this kind of a behaviour in which
mechanical deformation is coupled to electric phenomena. This phenomenon is called the
piezoelectric effect and the materials which display this behaviour are referred to as piezoelectric
materials. The piezoelectric effect was discovered in the 19th century by the Curie brothers. Their
experiments suggested that piezoelectric materials generate a potential difference when subject to
deformation. This behaviour is referred to as the direct piezoelectric effect and is shown in Fig. 3.1(a).
It follows that the opposite potential difference is generated when the opposite mechanical
deformation is imposed, i.e., the potential difference generated in tension is opposite to the potential
difference generated in compression. Further, these materials also exhibit the inverse piezoelectric
effect in which mechanical deformation of the material takes place depending on the potential
difference applied, as shown in Fig. 3.1(b). Over the past several decades the direct and inverse
piezoelectric effects have been used in applications like sonar, ultrasound and consumer electronics,
including smartphones.

a  b

Figure 3.1. Schematic of the (a) direct piezoelectric effect in which an applied deformation generates
a potential and the (b) inverse piezoelectric effect in which an applied potential generates elastic
deformation.

The Piezoelectric phenomenon

As with all physical phenomena, the piezoelectric behaviour observed for an object is driven by
the physical interactions that occur at the microscopic level, more specifically at the atomic level.
Piezoelectric materials are both naturally occurring and also synthetic, typical examples being quartz
for a natural piezoelectric material and ceramics such as Barium Titanate (BaTiO3) and Lead
Zirconate Titanate (PZT) for synthetic or manufactured piezoelectric materials. As mentioned
previously, the driver of the piezoelectric phenomenon occurs at the level of individual unit cells of
these materials. While the structure of the unit cell might be different for the natural and synthetic
piezoelectric materials, the operating principles and consequently, the constitutive relations and
problem descriptions remain same for both these materials. We now address the origins of the
piezoelectric behaviour with a specific emphasis on the crystal structure of synthetic piezoelectric
materials. The crystal structure of a synthetic piezoelectric material is shown in Fig. 3.2. One of the
key properties of these materials is that above a specific temperature, called the Curie temperature, the
crystal has a cubic lattice and a symmetric distribution of charges making it electrically neutral, as
shown in Fig. 3.2 (a). At temperatures below the Curie temperature, the positively charged metal ion
at the center moves slightly upward creating an orthorhombic crystal structure and more importantly
creating an imbalance in the charge distribution resulting in an electric dipole, as shown in Fig. 3.2
(b). The orientation of this electric dipole or the Polarization vector  P  is defined from the negative
charge to the positive charge. It is to be noted that the magnitude of P is a function of both the
magnitude of the charges in the electric dipole and the physical distance between them, i.e., an
increase in either of them results in an increase in the magnitude of P .

a  b

Figure 3.2. Crystal structure of a synthetic piezoelectric material (a) above the Curie temperature
displaying charge symmetry and (b) below the Curie temperature with a resultant electric dipole.

A typical piezoelectric ceramic is composed of possibly millions of grains, each of which is


composed of millions of polarized unit cells. Since the unit cells in one grain are oriented along the
same direction, each grain of the piezoelectric material exhibits polarity along the same direction as
the unit cell. However, since the piezoelectric material is composed of millions of grains, each of
which exhibits polarity in a random direction, the net polarity exhibited by a piezoelectric component
is essentially zero. This can be changed by subjecting the piezoelectric to a strong electric field which
reorients each grain in such a way that the polarization vector of that grain is oriented along the
applied electric field. The result is a piezoelectric component that exhibits polarity at the macroscopic/
component level.

a 
Direct effect

b
Inverse effect

Figure 3.3. (a) The direct piezoelectric effect which generates charges/ potential difference across a
piezoelectric material on deformation and (b) The inverse piezoelectric effect which generates
deformation on applying potential difference.
An example of such a piezoelectric material, which exhibits polarization, is shown in Fig. 3.3, to
the left. When this material is compressed by prescribing a stress or strain, then the individual unit
cells within the material also undergo compression and the unit cell shown in Fig. 3.2(b) becomes
smaller in the vertical direction. This effectively moves the positive charge density of the metal ions
downward and the negative charge density of the O atoms upward, resulting in an accumulation of
positive charge at the bottom of the cylinder and a negative charge at the top of the cylinder, as shown
in Fig. 3.3(a). Also, if a tensile stress/ strain is imposed instead of a compressive strain, then the
charge accumulation is the exact opposite with negative charges accumulating at the bottom and
positive charges accumulating at the top. Similarly, the inverse piezoelectric effect is explained using
the behaviour of a unit cell as follows. As shown in Fig. 3.3(b), if a potential is applied such that there
is an accumulation of negative charges at the top and positive charges at the bottom, then the
positively charged metal ions get attracted to the top surface and the negatively charged O atoms get
attracted to the bottom surface resulting in the expansion of the unit cell and consequently, the
piezoelectric component. Similar to the direct piezoelectric effect, imposing the exact opposite
conditions result in the opposite behaviour, i.e., applying a potential that accumulates positive charge
at the top and a negative charge at the bottom compresses the piezoelectric material.

Constitutive equations and BVP for piezoelectricity

We now address the constitutive equations for linear piezoelectricity bearing in mind that the
effects described in the previous section will have to be captured by the structure of the constitutive
equations. We first start by quantifying the fundamental behaviour demonstrated by the piezoelectric
material – the electric dipole moment or the polarization vector, P . We have previously stated that the
magnitude of the polarization vector is a function of both the magnitude of the charges forming the
dipole and the distance between them, i.e., any physical phenomenon that increases the distance
between the charges forming the dipole increases the magnitude of P . This is obviously
accomplished by the imposed stress as it serves to compress or expand the material, i.e., the
magnitude of the polarization vector should be dependent on the magnitude of stress imposed. While
not necessary, here we assume for simplicity that the polarization vector is linearly related to the
stress tensor as
P  d : (3.1)

Here, d is the third order electro-mechanical coupling tensor and governs the direction of the polarity
induced since the component form of Eqn. (3.1) reads Pi  dijk jk . This relation would determine the
behaviour of the piezoelectric material when an external stress/ strain is imposed, i.e., the direct
piezoelectric phenomenon. However, we note that the net electric parameters generated are not
described solely by the applied stress. For example consider the expansion of the piezoelectric
material under applied tensile stress as shown in Fig. 3.3(a). We have presented the argument that
when the piezoelectric material is stretched, it results in an accumulation of net positive charge at the
top and a net negative charge at the bottom of the specimen. These accumulated charges induce an
electric field directed from the top of the specimen to the bottom, i.e., the net electric charges/ fields
in the specimen are not only governed by the applied stress/ strain, but also by the electric properties
of the material. If the electric field due to the accumulation of the charges is represented by E , then
the strength of the electric field permeating inside the piezoelectric material is given by k  E , where
k is the second order tensor representing the electric permittivity of the material. Thus, any quantity
that describes the electric phenomenon of a piezoelectric material has to consider both the effect of
the applied stress and the induced electric field. We note here that the charges that develop the electric
field/ potential in the piezoelectric material are not free charges, but arise from the charges of the ions
in the material and are bound to the unit cells of the piezoelectric material. Thus, a more apt
description of the electric phenomenon can be obtained by using the Electric displacement vector
 D  , rather than the Electric field intensity. As mentioned before, this quantity should contain the
effect of both  and E , given by

D  k  E  d : (3.2)

In the above equation, we have formulated the effect of both the imposed stress and electric
properties of the material on the Electric displacement vector field, i.e., the above equation gives how
both the electric and mechanical phenomenon interact to produce a measurable electric parameter.
Similarly, we will have to quantify the interaction of the electric and mechanical phenomena to
produce a measurable mechanical parameter, say, strain. We know that when a material is subject to
an external stress, strain is generated. We can invert the linear elastic relation of Eqn. (1.8) to obtain
strain in terms of stress as   C1 :  where C is the fourth order stiffness tensor. We have accounted
for the direct piezoelectric behaviour in the previous equation, i.e., we have described the effect of
mechanical deformation on the electric displacement field created. Here, we address the inverse
piezoelectric behaviour, i.e., our aim will to quantify the effect of the generated electric field on the
deformation of the piezoelectric material. Referring to Fig. 3.3(b), we see that applying an electric
field in the same direction as the polarization vector elongates the specimen and an electric field
opposed to the polarization compresses the specimen. We know that when a piezoelectric material is
acted upon by an external stress, it compresses or elongates and accumulates charges on the boundary
which creates an electric field. This electric field in turn generates additional strain in the piezoelectric
material due to the inverse piezoelectric effect. Thus, the total strain in a piezoelectric material is a
combination of the strain due to applied stress and due to the inverse piezoelectric effect. This
additional strain generated due to an electric field is described using the strength of the electric field,
E and d , the third order electro-mechanical coupling tensor. The total strain is then given as

  C1 :   E  d (3.3)

The above relations describe the constitutive behaviour of piezoelectric materials. Both Eqns. (3.2
and 3) reflect the strongly coupled nature of piezoelectricity, i.e., the electric displacement is
dependent on both the electric field intensity and the stress imposed and similarly, the strain
undergone by the specimen depends on the applied stress and the electric field intensity. To describe
the complete boundary value problem, we also need to formulate the governing equations and the
boundary conditions. A schematic of the BVP for the direct piezoelectric effect is shown in Fig. 3.4
with points A and B on the boundary constrained and traction t applied over part of the boundary, t .
In addition, the electric potential over part of the boundary,  is prescribed to be 0. This specific
condition should not be necessary in the strictest mathematical sense since an electric field being
generated implies a unique potential difference rather than the value of the electric potential itself.
However, numerical methods used to solve for the boundary value problems usually converge to
solutions when such a boundary condition is included. We do realize that the primary variables of this
electro-elastic boundary value problem are the electric potential and the displacement vector.

0 Figure 3.4. Schematic of the boundary


0 value problem for describing the direct
B t
ê3 piezoelectric effect.

A
ê1 ê2

For the direct piezoelectric phenomenon, we would have two governing equations since we have two
sets of physics interacting with each other – the electric phenomenon and the mechanical
phenomenon. We know that the mechanical phenomenon is governed by the balance of linear
momentum which we have also used previously. To describe the electric phenomenon, we have
previously used the conservation of charge expressed as the divergence of the electric field. In the
present situation, we have reasoned that using the electric displacement vector is more relevant than
using the electric field strength vector since the electric phenomenon includes the effect of the electric
dipole moment, which is being generated by bound charges, i.e., if the total charge density is the sum
of the free and the bound charges,

   free  bound (3.4)

Here the density of the bound charges is related to the electric dipole moment (Polarization) vector as
bound    P and the net charge density is related to the electric field strength as

   k  E   (3.5)

Then, from the expression relating the bound charge density and the electric dipole moment and Eqn.
(3.4), we can write,

   k  E    P    free    D   free (3.6)

Given that the electric phenomenon is governed by the charges bound to a unit cell and there is no
free flowing electric current, i.e.,  free  0 , we can simplify Eqn. (3.6) to write the governing
equation for the electric phenomenon as
D  0 (3.7)

We know that the governing equation for the linear elastic mechanical deformation is given by the
balance of momentum and for the time independent case, it is given by
   b  0 (3.8)

The above equations, Eqns. (3.7 and 8) are the governing differential equations for the piezoelectric
phenomenon. We however note that the constitutive relations are not consistent with the GDEs, i.e.,
Eqn. (3.3) describes the strain evolution, whereas Eqn. (3.8) is in terms of the stress. We can however,
rewrite Eqn. (3.3) in terms of stress as,


C1 :     E  d    C :   CT : dT  E (3.9)

Also, we can write Eqn. (3.2) as


D  k  E  d :C:  D  k E  e: (3.10)

Here, e  d : C is defined as the third order tensor of Piezoelectric constants. It follows that
eT  CT : dT using which Eqn. (3.9) can be rewritten as

  C :   eT  E (3.11)

We can now write the boundary value problem for piezoelectric phenomenon as

    b  0
 Governing equations.
D  0 

  C :   eT  E 
 Constitutive relations (3.12)
D  k  E  e :  

u A  u B  0, nˆ    t on t 
 Boundary conditions.
  0 on 0 

1
Of course, here E   and    u  u  . The above form of the constitutive equations given in
2
Eqn. (3.12) is called as the Stress-Charge form of the piezoelectric equations.

We have introduced a new tensor here, the piezoelectric tensor, given by e  d : C . Since d is a
third order tensor and C is a fourth order tensor, e would be a third order tensor which would have
three indices and 27 components. However, similar to the stiffness tensor, most of the components are
0 due to the symmetry. We can write the constitutive relations given in Eqn. (3.12) in indicial form as

Di  kij E j  eijk  jk
(3.13)
 ij  Cijkl  kl  emij Em

We see in the above equation that the second and third indices of the e tensor always act together.
Similar to stress and strain tensors for which it has been defined, we can use the Voigt notation to
represent the components of the e tensor as well, except that we restrict this to the second and third
indices of the e tensor . We can do this since the symmetry of the stress tensor requires that
 ij   ji  emij Em  emji Em  emij  emji . Similar to the Voigt notation we can club the last two
indices of the e tensor using a single index, using the symmetry property, i.e., eijk  eiJ , where i, j
and k can take values 1, 2 and 3, but J takes values between 1 and 6. Then the relations of Eqn. (3.13)
can then be written as
Di  kij E j  eiJ  J
(3.14)
 I  C IJ  J  emJ Em

The matrix representation of the above quantities will be as follows. The electric displacement vector
 D  is a vector and is represented as a 3X1 matrix, the permittivity tensor  k  is a second order tensor
and is represented as a 3X3 matrix and the electric field strength  E  is a vector field represented as a
3X1 matrix. The stress and stain tensors are second order tensors and can be represented as 3X3
matrices. In the above representation, since we have incorporated the Voigt notation, these quantities
would be represented as 6X1 matrices. Also, the fourth order stiffness tensor can be represented as a
6X6 matrix using the Voigt notation as addressed in earlier chapters. The third order tensor of
piezoelectric properties would have been a three dimensional data structure with 27 components.
However, due to symmetry considerations and the use of the Voigt notation, we can reduce the
number of indices required to two as shown in Eqn. (3.14), and can express it in a matrix form. Using
the Voigt notation, we have expressed the components of the e tensor as e  eijk  eiJ where i can
take values of 1, 2 and 3 and J can take values from 1 to 6. Thus, the matrix representation of e will
be of 3X6 size. Written out in terms of the size of the matrices used to represent the relevant terms,
the constitutive relations are expressed as

D 31  k 33 E31  e36   61


(3.15)
  61  C 66   61  eT 63 E31

In the direct piezoelectric phenomenon, we see that the e tensor interacts with the components of
the small strain tensor to produce the electric displacement vector. For example, the e 25 element in the
e matrix is multiplied with  5 to contribute to the electric displacement vector along the ê 2 direction.
In reality, we know that it is actually the e213 component of the e tensor interacting with 13 . An
example of the components of the e tensor expressed in a matrix form using the Voigt notation for
e11 e12 0 e14 0 0 

quartz is  e   0 0 0 0 e25 e26  where e12  e26  e11 and e25  e14 . We see from the
 0 0 0 0 0 0 
matrix representation that not only normal components of deformation, but shear deformations also
induce the piezoelectric effect. We also see that there is no deformation that can be prescribed that
would generate polarization along the ê3 direction. The sense of the eˆ1 ,eˆ 2 and eˆ 3 directions is defined
along the directions of the unit crystal cell. We note here that the identity of the components of the e
matrix that are non-zero are dependent on the material itself, i.e., the e matrix for a material such as
BaTiO3 or a PZT ceramic could be completely different.

You might also like