Chapter 2

Indicial Relations and Launching Platform

My father held in a very responsible government position and he used to call his stenographer
often to our home on weekends. As a child I was amazed to find that my father would dictate
letters after letters at a conversational pace and the stenographer would have no problems
with the speed. Then I asked the secret from the stenographer. He had taken a training to
write in shorthand and he absolutely had no problem in noting down the words spoken by
my father.

In the field of solid mechanics, we deal with many equations, sometimes 20 equations or so.
And many of them are with partial differentials. We, therefore, have also evolved a concise
way of writing the equations. An indicial relation reduces the number of equations; from three
to one for a vector equation and nine to one for a second order tensorial equation. Further,
Einstein summation convention reduces the length of an equation considerably. In this
chapter, we will be learning how to deal with indicial equations and how to use Einstein
summation convention. Then in the entire book we will play with the indicial relations. Also,
we will develop the required mathematical background in light of the indicial notations.

2.1 Space Variables and Coordinate System

If we want to do stress analysis of a crankshaft, we may like to find the point at which the
effect of stress tensor is most severe. Such a point is known as critical point. We analyze the
entire body of the crankshaft to find the critical point. Thus, we move in all the three space
directions. Most commonly used coordinate system is Cartesian with axes as x, y, and z. But
for indicial equations, it is convenient to deal with numbers and, therefore, we will use space
coordinate axes as x1, x2, x3 as shown in Fig. 2.1a. We make sure that the axes orientation is
chosen following the right-hand rule. Some problems are convenient to solve in the cylindrical
coordinate system r, 𝜃, z as shown in Fig. 2.1b. Unlike in Cartesian coordinate system, in this
coordinate system we usually do not address them through numbers. The angle 𝜃 is measured
from a reference line on a plane which is normal to z-axis. The universally accepted
convention is that r comes first, 𝜃 next and, then z. There are some more coordinate systems
such as spherical coordinate system. We will discuss them when a need arises.
 Fig. 2.1 (a) Cartesian coordinate system, and (b) cylindrical coordinate system

Some people prefer to deal with symbols which are not tied to any coordinate system, such
as a vector is represented just by v instead of (v1, v2, v3). This approach is more general and
powerful, and often more compact also. In this book we will try to follow both approaches so
that you will be able to follow research papers in journals. The indicial form is only for
Cartesian coordinate system but often gives better feel of a problem.

The commonly used dependent variables in solid mechanics are stress σ, strain ϵ, or
displacement u. We can choose any one to solve a problem. The choice usually depends on
the nature of the problem and the associated boundary conditions. If we solve for
displacement u, we can easily determine strain components by differentiating the
displacement components with space variables. For most materials, stress-strain relations are
available and using them we can determine stress components. On the other hand, if a
problem is solved for stress components, strains are determined using stress-strain relations.
If needed, strain components thus obtained can be integrated to obtain displacement
components.

In some problems, there may be more involved dependent variables such as buckling load of
a column, strain energy density, stress concentration, stress intensity factor in fracture
mechanics, von Mises stress, etc. We will not bother about them at this stage. Whenever a
new kind of dependent variable is involved, it will be defined and developed right there.

2.2 Continuum

Partial differential equations are extensively used in Solid mechanics. Thus, the variables
should be continuous from a point to another so that we can use calculus. In case there exists
a discontinuity, especial attention is required in such a way that the equations are not invoked
across the discontinuity.
It is well known that a material is made of atoms. Thus, there exists gaps between
neighbouring atoms. In a rigorous sense, discontinuity exists from one point to another as
shown in Fig. 2.2a. The gaps between the atoms is of the order of only several Angstroms.
Since 1 angstrom is equal to 0.1 nm, the gap between atoms is very small, less than a nm. The
dimensions of most structural members are far greater than the gap between the atoms. We
can then consider a real-life member to be continuous for independent space variables x1, x2,
x3. However, if there are holes and cuts within the material of a member, one cannot use the
field equations across the cuts, say from point A to point B directly in Fig. 2.2b. Such problems
can be solved by staying within the continuous material; for example, we go from point A to
Point C and then to point B. Since only continuous members can be solved through the field
equations, some people like to call it continuum mechanics.

A caution must be mentioned here. If we are dealing with nanostructures such as a very thin
film of a few atomic thick, the equations we will be developing in this book are not relevant
anymore. The domain is different, and we may develop equations based on atomistic model.
One should always remember the domain within which the analysis is developed and solved.
In the modern world, we are developing many new kinds of materials and creating various
new products. Sometimes we do not realize that a new problem is quite different and does
not fall within the domain we are familiar with.

Fig.2.2 (a) Gaps between atoms, and (b) a member with cuts

2.3 Scalars, vectors and tensors

Mathematical variables are of various kinds, scalars, vectors and tensors.

Scalars: The simplest ones are scalars which are not associated with any space direction. They
are many such variables but some of them for our interest are: time, mass, volume, density,
temperature, various kinds of energies, entropy, work. Each one of them is fully described
just by one number.

Vectors: A vector is slightly is more complex and is associated with a direction. Since a
direction is defined by three space variables, a vector 𝒗 is described with three components,
𝑣1 , 𝑣2 , 𝑣3 . If e1, e2, e3 are three unit-vectors in direction x1, x2, x3, vector 𝒗 can be expressed
as (Fig 2.1):

𝒗 = 𝑣1 𝒆𝟏 + 𝑣2 𝒆𝟐 + 𝑣3 𝒆𝟑 … (2.1)

The magnitude of the vector is |𝒗|or just 𝑣 is:

𝑣 = √(𝑣12 + 𝑣22 + 𝑣32 ) … (2.2)

The vector v is also expressed as:

𝒗 = (𝑣1 , 𝑣2 , 𝑣3 ) … (2.3)

We can also express vector 𝒗 as 𝑣𝑖 where subscript, 𝑖, is known as free index. It can take 1, 2,
or 3 value. In other words, if we prescribe 𝑖 =1, it refers to the first component of vector 𝑣1 ,
𝑖 =2 refers to 𝑣2 , and 𝑖 =3 to 𝑣3 .

𝑣 = 𝑣𝑖 … (2.4)

It is worth noting that Eq. 2.3 is simpler than Eq. 2.1, and Eq. 2.4 is made more compact
further. This form will be exploited extensively in developing field equations of solid
mechanics.

There are many vector variables of importance to us. Some of them are distance, velocity,
acceleration, angular acceleration, force, momentum, moment, torque, and angular
momentum.

Tensors: A relation between two vectors depends on two directions, one of each vector. The
relation is more complex as it has to manage two directions. If 𝒑 and 𝒒 are two vectors, the
relation, in general, can be quite complex. But the simplest one is a linear relation which is
expressed as:

𝒑 = 𝑆𝒒 … (2.5)

where 𝑆 is a second order tensor. The equation can be expressed in component-form as:

𝑝1 = 𝑆11 𝑞1 + 𝑆12 𝑞2 + 𝑆13 𝑞3

𝑝2 = 𝑆21 𝑞1 + 𝑆22 𝑞2 + 𝑆23 𝑞3 … (2.6)

𝑝3 = 𝑆31 𝑞1 + 𝑆32 𝑞2 + 𝑆33 𝑞3

Alternatively, these relations can be expressed in a matrix form as:

𝑝1 𝑆11 𝑆12 𝑆13 𝑞1

𝑝
{ 2 } = [𝑆21 𝑆22 𝑆23 ] {𝑞2 } … (2.7)
𝑝3 𝑆31 𝑆32 𝑆33 𝑞3
In the above equation, the second order tensor 𝑆 keeps track of the direction of vector 𝒑 as
well as that of vector 𝒒. Therefore, 𝑆 has two indices representing three components of vector
𝒑 and three components of vector 𝒒, resulting into nine components of tensor S.

The second order tensor 𝑆 can be represented in several ways; (i) just 𝑆, (ii) through a 3x3
matrix, or (iii) just 𝑆𝑖𝑗 . In 𝑆𝑖𝑗 , 𝑖 and 𝑗 are free indices, each can take value from 1 to 3. For
example, if 𝑖 = 2 and 𝑗 = 3, the component of 𝑆 is 𝑆23 . We thus find that 𝑆𝑖𝑗 is a 3x3 matrix
and is quite compact – a very useful feature. In later chapters, we will be writing the second
order stress tensor 𝜎 as 𝜎𝑖𝑗 , and second order strain tensor 𝜖 as 𝜖𝑖𝑗 .

We also come across third order tensors in our real life. For example, one of my PhD students
designed a piezoelectric sensor where the electric voltage 𝒗, a vector in a single crystal,
depended on the stress tensor 𝜎 with the relation:

𝒗 = (𝑑)𝜎 … (2.8)

where 𝑑 is a third order tensor accounting two directions of stress tensor and one direction
of voltage. The third order tensor has 3x3x3=27 components. We can as well express it in
component form as 𝑑𝑖𝑗𝑘 where 𝑖, 𝑗, 𝑘 are free indices.

Do we deal with higher order tensors? Yes, we do. A fourth order tensor is extensively used
in solid mechanics. The property of a material relates second order stress tensor with second
order strain tensors by the relation:

𝜎 = 𝐶𝜖 … (2.9)

where 𝐶 is a fourth order tensor. It accounts for two directions of stress tensor 𝜎 and two
directions of strain 𝜖. It has 34 = 81 components and indicial form we address it as 𝐶𝑖𝑗𝑘𝑙 .

When we look back and try to take a bird eye view, we find that tensor is a convenient and
compact form of keeping track of complex equations. Tensor was first developed by
______________in ______. The tensorial relations are found to be convenient in various
computer software packages. In fact, a vector which is associated with only one direction is a
tensor of first order. We, however, do not call vectors to be first order tensor because it was
developed much earlier than the formulation of tensor. I give you an example here.
Historically, we realized the importance of half for ½ quite early in the development of human
understanding. However, one-third, one fourth, etc. were realized much later. And if we do
call half as one-second, it sounds odd, although rigorously speaking one-second makes more
sense. No body now will dare to call vectors as the first order tensors.
 2.4 Indicial Relations

Consider a simple relation:

𝑏𝑖 = 𝑎𝑐𝑖 + 𝑑𝑖 … (2.10)

In this equation, 𝑏𝑖 , 𝑐𝑖 and 𝑑𝑖 are vectors, 𝑎 is a constant, and 𝑖 is the free index. For 𝑖 = 1,
we obtain 𝑏1 = 𝑎𝑐1 + 𝑑1 and similarly two other equations for 𝑖 = 2 and 𝑖 = 3. Thus, Eq. 2.10
represents the following three relations,

𝑏1 = 𝑎𝑐1 + 𝑑1

𝑏2 = 𝑎𝑐2 + 𝑑2 … (2.11)

𝑏3 = 𝑎𝑐3 + 𝑑3

In this particular case, the indicial relation of Eq. 2.10 is a convenient way of representing 3
equations by only one relation. This makes the language of mathematics more compact; three
equations are compacted into one without compromising on rigor. Finding a new way of
representing something is creativity of high order. You may not realize that ‘=’ symbol was
formulated by Welsh mathematician Robert Recorde in 1557. Before this symbol was
invented, people used to write mathematical relations in long form.

To understand the magic of indicial equations, consider an equation having a second order
tensorial terms:

𝜕𝑐
𝐵𝑖𝑗 = 𝑚 𝜕𝑥𝑖 + 𝐷𝑗𝑖 … (2.12)
𝑗

In this equation, 𝐵 and 𝐷 are second order tensors, 𝑚 a constant, and 𝑐𝑖 is a vector which is
partially differentiated by space variable 𝑥𝑗 . For 𝑖 = 1 and 𝑗 = 3, the corresponding equation
is:

𝜕𝑐
𝐵13 = 𝑚 𝜕𝑥1 + 𝐷31 … (2.13)
3

Eq. 2.12 compresses 9 equations into one because 𝑖 can take 3 values and 𝑗 can also take 3
values. In solid mechanics, there are cases when we deal with a large number of equations,
sometimes may be of the order of 24, to solve a problem. Using indicial form, they are
compressed to only 5 equations. We tend to get lost in the jungle of 24 equations while 5
equations are like the backyard of a home. Such compactness allows us to see deeper into
the field of solid mechanics. Also, it saves time and drudgery of writing many equations. In
the next section, we will discuss the Einstein summation convention which reduced the length
of an indicial equation, compacting the equations further.

Example 2.1: Express the following three equations through an indicial equation:
 𝜕𝑔
𝑢1 = 𝑎1 𝑏 + 𝜕𝑥
1

𝜕𝑔
𝑢2 = 𝑎2 𝑏 + 𝜕𝑥 … (2.14)
2

𝜕𝑔
𝑢3 = 𝑎3 𝑏 + 𝜕𝑥
3

where 𝑔 is a function of 𝑥1 , 𝑥1 , and 𝑥3 .

Solution: In all the three equations, vector 𝑎 takes three values 𝑎1 , 𝑎1 , and 𝑎3 and partial
𝜕 𝜕 𝜕
differential also take three values , and . We combine them into one indicial
𝜕𝑥1 𝜕𝑥2 𝜕𝑥3
equation as:

𝜕𝑔
𝑢𝑖 = 𝑎𝑖 𝑏 + 𝜕𝑥 … (2.15)
𝑖

Example 2.2: Express the following equations through one indicial equation:
𝐷11 = 𝑎𝐵11 + 𝑐1 ; 𝐷12 = 𝑎𝐵12 + 𝑐1 ; 𝐷13 = 𝑎𝐵13 + 𝑐1
𝐷21 = 𝑎𝐵21 + 𝑐2 ; 𝐷13 = 𝑎𝐵13 + 𝑐2 ; 𝐷23 = 𝑎𝐵23 + 𝑐2
𝐷31 = 𝑎𝐵31 + 𝑐3 ; 𝐷32 = 𝑎𝐵32 + 𝑐3 ; 𝐷33 = 𝑎𝐵33 + 𝑐3

Solution: The indices of tensor 𝐷 and 𝐵 are same and the first index of 𝐷 is same as that of
vector 𝒄. This leads to indicial equation as:

𝐷𝑖𝑗 = 𝑎𝐵𝑖𝑗 + 𝑐𝑖

2.5 Einstein Summation Convention

The famous scientist Albert Einstein introduced the summation convention in 1916. We find
it very useful and convenient and use it extensively in solid mechanics.
Consider an equation:

𝑠 = 𝑎1 𝑥1 + 𝑎2 𝑥2 + 𝑎3 𝑥3 … (2.16)

It is usually expressed as:

𝑠 = ∑3𝑖=1 𝑎𝑖 𝑥𝑖 … (2.17)

Einstein suggested that we need not write ∑3𝑖=1 and simplify the equation to:

𝑠 = 𝑎𝑖 𝑥 𝑖 … (2.18)
In Solid Mechanics, we usually sum with respect to space variable 𝑥1 , 𝑥1 , and 𝑥3 . Thus, 𝑖
varies from 1 to 3. The Einstein convention conveniently shrinks the length of an equation.
In the equation, 𝑖 is known as dummy index, and its symbol can be changed if needed.

Therefore, Eq.2.18 can be written as

𝑠 = 𝑎𝑗 𝑥𝑗
or, 𝑠 = 𝑎𝑘 𝑥 𝑘

We do require to change the symbol once in a while in solid mechanics.

There are several important aspects of dummy index:

(i) A dummy index is confined within a term

(ii) It occurs only twice in a term. For example, we cannot write 𝑎𝑖 𝑏𝑖 𝑐𝑖 𝑑𝑖 . We can have
two dummy indices as 𝑎𝑖 𝑏𝑖 𝑐𝑗 𝑑𝑗

An equation may have both kinds of indices, free indices and dummy indices. let us consider
an example as:

𝑡𝑖 = 𝜎𝑖𝑗 𝑛𝑗 … (2.19)

In this relation, 𝑡𝑖 and 𝑛𝑗 are vectors and 𝜎𝑖𝑗 is a second order tensor. It represents three
equations as:
𝑡1 = 𝜎11 𝑛1 + 𝜎12 𝑛2 + 𝜎13 𝑛3
𝑡2 = 𝜎21 𝑛1 + 𝜎22 𝑛2 + 𝜎23 𝑛3 … (2.20)

𝑡3 = 𝜎31 𝑛1 + 𝜎32 𝑛2 + 𝜎33 𝑛3

In this example, 𝑖 is the free index and 𝑗 the dummy index. It is worth noting that Eq. 2.19
represents a vector equation because there are three components of vector 𝒕 . A free index
occurs at the most only once per term in an indicial relation. Just to refresh your knowledge,
terms in an equation are separated by + or – sign. For example, there are five terms in the
following equation:
𝑢𝑖 = 𝑆𝑚𝑖 𝑣𝑚 − 𝐴𝑖𝑘 𝐵𝑘𝑙 𝐶𝑙 + 𝑑𝑖 − 𝑝 … (2.21)
In this equation 𝑖 is the free index 𝑚, 𝑘, 𝑙 are dummy indices and 𝑝 is a constant. Note that 𝑖
appears once in the first four terms and does not appear in the last term. It is a vectorial
relation and it represents three equations. Also note that a dummy indices are confined
within a term and there are two dummy indices in the third term. We now consider an
equation where there are two free indices:

𝑇𝑖𝑗 = 𝑆𝑖𝑘 𝐵𝑘𝑗 + 𝐴𝑙𝑖 𝐵𝑗𝑙 … (2.22)

In this relation, 𝑇, 𝑆, 𝐴, and 𝐵 are second order tensors, each represented by two indices. It
consists of two free indices 𝑖 and 𝑗, and two dummy indices 𝑘 and 𝑙. This is a relation of second
order tensors and represents 9 equations.
When a second order tensor operates on a vector it yields a vector. For example, in equation
𝒖 = 𝑆𝒗, second order tensor 𝑆 acts on vector 𝒗 to yield vector 𝒖. In component form it is
written as:
𝑢𝑖 = 𝑆𝑖𝑗 𝑣𝑗 … (2.23)

There is only one free index 𝑖 and one dummy index 𝑗.

When a second order tensor acts on another second order tensor, it yields a second order
tensor. In relation 𝐹 = 𝑄𝑈, second order tensor 𝑄 act on another second order tensor 𝑈 to
yield a second order tensor 𝐹. In indicial form, the equation is written with the help of a
dummy index 𝑘 as:

𝐹𝑖𝑗 = 𝑄𝑖𝑘 𝑈𝑘𝑗 … (2.24)

On similar lines, a fourth order tensor 𝐶 acting on a second order tensor 𝜖 yields a second
order tensor 𝜎 as 𝜎 = 𝐶𝜖. In component form it is expressed as:

𝜎𝑖𝑗 = 𝐶𝑖𝑗𝑘𝑙 𝜖𝑘𝑙 … (2.25)

Although a fourth order tensor 𝐶𝑖𝑗𝑘𝑙 is involved, the equation is of second order tensor
because there are only two free indices and it represents 9 equations.
An equation may have vectors and tensors in its expression, but it may still be a scalar
equation if it does not have free indices. For example, consider the equation:
𝑠 = 𝐴𝑖𝑗 𝐵𝑗𝑖 + 𝑢𝑘 𝑣𝑘 + 𝑐 … (2.26)

In this equation, 𝐴 and 𝐵 are second order tensors, 𝒖 and 𝒗 are vectors, and 𝑖, 𝑗 and 𝑘 are
dummy indices. But it has no free index and, therefore, it represents only one equation and
thus 𝑠 is a scalar.
In short, no free index in an equation means it is a scalar equation, one free index in a relation
represents a vector equation representing 3 equations, and an equation with two free indices
corresponds to a second order tensorial relation representing 9 equations. Thus, it pays to
keep track of number of free indices in an indicial equation. Also, in a term a vector or tensor
acts on what is present to the right of it. Consider the equation:
𝒖 = 𝑄𝑈𝒗 … (2.27)
The second order tensor 𝑈 acts on vector 𝒗 to yield a vector. Then another second order
tensor 𝑄 acts on the vector 𝑈𝒗 to yield vector 𝒖.
Example 2.3: Write the following in full form: (a) 𝐴𝑖𝑖 ; (b) 𝐶𝑗𝑗 𝐷𝑘𝑘 ; (c) 𝑐𝑖 𝐷𝑘𝑘 ;

(d) 𝐷𝑖𝑘 = 𝐴𝑖𝑗 𝐵𝑗𝑘

Solution:

a) Since 𝑖 is a dummy index, 𝐴𝑖𝑖 = 𝐴11 + 𝐴22 + 𝐴33

b) This expression is also a scalar as 𝑗 and 𝑘 are dummy indices. Thus,
𝐶𝑗𝑗 𝐷𝑘𝑘 = (𝐶11 + 𝐶22 + 𝐶33 ) × (𝐷11 + 𝐷22 + 𝐷33 )
c) 𝑖 is free index and 𝑘 is dummy index, yielding:
𝑐𝑖 𝐷𝑘𝑘 = { 𝑐1 (𝐷11 + 𝐷22 + 𝐷33 ) 𝑐2 (𝐷11 + 𝐷22 + 𝐷33 ) 𝑐3 (𝐷11 + 𝐷22 + 𝐷33 )}
d) 𝑖 and 𝑘 are free indices and𝑗 is dummy index and thus it represents 9 equations:
𝐷11 = 𝐴11 𝐵11 + 𝐴12 𝐵21 + 𝐴13 𝐵31 ; 𝐷12 = 𝐴11 𝐵12 + 𝐴12 𝐵22 + 𝐴13 𝐵32
𝐷13 = 𝐴11 𝐵13 + 𝐴12 𝐵23 + 𝐴13 𝐵33 ; 𝐷21 = 𝐴21 𝐵11 + 𝐴22 𝐵21 + 𝐴23 𝐵31
𝐷22 = 𝐴21 𝐵12 + 𝐴22 𝐵22 + 𝐴23 𝐵32 ; 𝐷23 = 𝐴21 𝐵13 + 𝐴22 𝐵23 + 𝐴23 𝐵33
𝐷31 = 𝐴31 𝐵11 + 𝐴32 𝐵21 + 𝐴33 𝐵31 ; 𝐷32 = 𝐴31 𝐵12 + 𝐴32 𝐵22 + 𝐴33 𝐵32

𝐷33 = 𝐴31 𝐵13 + 𝐴32 𝐵23 + 𝐴33 𝐵33

Example 2.4: Express the following equation through an indicial equation:

𝜕𝜎11 𝜕𝜎12 𝜕𝜎13
+ + + 𝑓1 = 0
𝜕𝑥1 𝜕𝑥2 𝜕𝑥3

𝜕𝜎21 𝜕𝜎22 𝜕𝜎23

+ + + 𝑓2 = 0 … (2.28)
𝜕𝑥1 𝜕𝑥2 𝜕𝑥3

𝜕𝜎31 𝜕𝜎32 𝜕𝜎33

+ + + 𝑓3 = 0
𝜕𝑥1 𝜕𝑥2 𝜕𝑥3

In these equations, 𝜎 is a second order tensor and 𝒇 is a vector.

Solution: These equations are vectoral relation because there is only one free index, first index
of 𝜎 and index of vector 𝒇. The second index of 𝜎 is same as that of 𝑥 in the same term,
indicating the existence of a dummy index. Thus, the indicial equation becomes:

𝜕𝜎𝑖𝑗
+ 𝑓𝑖 = 0
𝜕𝑥𝑗

𝟐. 𝟔 𝐊𝐫𝐨𝐧𝐞𝐜𝐤𝐞𝐫 𝐃𝐞𝐥𝐭𝐚
Kroneckar delta, 𝛿𝑖𝑗 , is a double indicial variable and plays important roles in solid mechanics.
It was created by German mathematician Leopold Kroneckar (1823-1891) and is defined as:
𝛿𝑖𝑗 = 1 if 𝑖 = 𝑗
𝛿𝑖𝑗 = 0 if 𝑖 ≠ 𝑗
In nonindicial form, it is called identity matrix 𝐼 expressed as:
1 0 0
𝐼 = 0 1 0] = 𝛿𝑖𝑗
[
0 0 1
Example 2.5: Evaluate: (a) 𝛿𝑖𝑖 ; (b) 𝛿𝑖𝑖 𝛿𝑘𝑘 ; (c) 𝛿𝑖𝑗 𝛿𝑖𝑗 ; (d) 𝛿𝑖𝑗 𝛿𝑗𝑘 𝛿𝑘𝑖
Solution:

a) 𝛿𝑖𝑖 : 𝑖 is the dummy index which means summation. Thus,

 𝛿𝑖𝑖 = 𝛿11 + 𝛿22 + 𝛿33 = 1 + 1 + 1 = 3
b) 𝛿𝑖𝑖 𝛿𝑘𝑘 =( 𝛿11 + 𝛿22 + 𝛿33 ) × (𝛿11 + 𝛿22 + 𝛿33 ) = 3 × 3 = 9
c) 𝛿𝑖𝑗 𝛿𝑖𝑗 : Since 𝑖 is equal to 𝑗, we invoke the definition in second 𝛿𝑖𝑗 on the first one to
have:
𝛿𝑖𝑗 𝛿𝑖𝑗 = 𝛿𝑖𝑖 = 3
d) 𝛿𝑖𝑗 𝛿𝑗𝑘 𝛿𝑘𝑖 =𝛿𝑖𝑗 𝛿𝑗𝑖 = 𝛿𝑖𝑖 = 3

Example 2.6: Simplify: (a) 𝛿𝑖𝑗 𝛿𝑗𝑘 ; (b) 𝑣𝑖 𝛿𝑖𝑗 ; (c) 𝐴𝑖𝑗 𝛿𝑖𝑘 𝛿𝑗𝑙 ; (d) (𝐴𝑖𝑗 − 𝐴𝑗𝑖 )𝛿𝑖𝑗 ;

(e) 𝐴𝑖𝑗 𝐵𝑗𝑘 𝛿𝑖𝑚 𝛿𝑘𝑛

Solution:

a) 𝛿𝑖𝑗 𝛿𝑗𝑘 : 𝑖 and 𝑘 are free indices and 𝑗 dummy index. We take the dummy index out to
have: 𝛿𝑖𝑗 𝛿𝑗𝑘 = 𝛿𝑖𝑘
b) 𝑣𝑖 𝛿𝑖𝑗 : In this expression, 𝑖 is the dummy index and j is the free index. 𝛿𝑖𝑗 states that
value of vector 𝒗 is nonzero only when 𝑖 = 𝑗. Thus, 𝑣𝑖 𝛿𝑖𝑗 = 𝑣𝑗
c) 𝐴𝑖𝑗 𝛿𝑖𝑘 𝛿𝑗𝑙 : In this one, 𝑖 and 𝑗 are dummy indices and it is simplified to 𝐴𝑘𝑙
d) (𝐴𝑖𝑗 − 𝐴𝑗𝑖 )𝛿𝑖𝑗 = (𝐴𝑖𝑖 − 𝐴𝑖𝑖 ) = 0
e) 𝐴𝑖𝑗 𝐵𝑗𝑘 𝛿𝑖𝑚 𝛿𝑘𝑛 : 𝑚 and 𝑛 are free indices, and 𝑖 and 𝑘 are dummy indices simplifying to
𝐴𝑚𝑗 𝐵𝑗𝑛 .

Example 2.7: Express following nine equations into one indicial equation:
𝜎11 = 𝜆(𝜖11 + 𝜖22 + 𝜖33 ) + 2𝜇𝜖11
𝜎22 = 𝜆(𝜖11 + 𝜖22 + 𝜖33 ) + 2𝜇𝜖22
𝜎33 = 𝜆(𝜖11 + 𝜖22 + 𝜖33 ) + 2𝜇𝜖33
𝜎12 = 2𝜇𝜖12
𝜎13 = 2𝜇𝜖13
𝜎21 = 2𝜇𝜖21
𝜎23 = 2𝜇𝜖23
𝜎31 = 2𝜇𝜖31
𝜎32 = 2𝜇𝜖32

Solution: In the last six equations, indices of 𝜎 are same as those of 𝜖. In the first three
equations also, the indices of 𝜎 on the left-side are same as those of the last term. However,
𝜆(𝜖11 + 𝜖22 + 𝜖33 ) is missing in the last six equations. Further, it can be expressed as 𝜖𝑘𝑘
using the summation convention. Now we can make a good use of Kronecker delta to have
the indicial equation as:

𝜎𝑖𝑗 = 𝜆𝛿𝑖𝑗 𝜖𝑘𝑘 + 2𝜇𝜖22

Comments/Inference: very compact indicial equation expressing nine equations with only
one equation and it is shorter in length too.

Example 2.8: Obtain only one equation in terms of components of vector 𝒖.

 𝜎𝑖𝑗,𝑗 + 𝐹𝑖 = 0 … (2.28 a)

𝜎𝑖𝑗 = 𝜆𝛿𝑖𝑗 𝜖𝑘𝑘 + 2𝜇𝜖𝑖𝑗 … (2.28 b)

1
𝜖𝑖𝑗 = 2 (𝑢𝑖,𝑗 + 𝑢𝑗,𝑖) ) … (2.28 c)

Strategy: We substitute Eq. (c) into Eq. (b) to eliminate 𝜖. Then we substitute the resulting
equation into Eq. (a).

Solution: In Eq. (c) we obtain 𝜖𝑖𝑖 by replacing 𝑗 by 𝑖 and then the dummy index symbol is
changed to obtain 𝜖𝑘𝑘 needed in Eq. (b). We then have:
1
𝜖𝑘𝑘 = 2 (𝑢𝑘,𝑘 + 𝑢𝑘,𝑘 ) = 𝑢𝑘,𝑘

Substituting in Eq. (b), we obtain:

1
𝜎𝑖𝑗 = 𝜆𝛿𝑖𝑗 𝑢𝑘,𝑘 + 2𝜇[2 (𝑢𝑖,𝑗 + 𝑢𝑗,𝑖) )]

Differentiating it to obtain:

𝜎𝑖𝑗,𝑗 = 𝜆𝛿𝑖𝑗 𝑢𝑘,𝑘𝑗 + 𝜇(𝑢𝑖,𝑗𝑗 + 𝑢𝑗,𝑖𝑗 )

Invoking definition of 𝛿𝑖𝑗 , we have:

𝜎𝑖𝑗,𝑗 = 𝜆𝑢𝑘,𝑘𝑖 + 𝜇(𝑢𝑖,𝑗𝑗 + 𝑢𝑗,𝑖𝑗 )

Replacing dummy index j of the last term by k and then adding it to the first term on right-
side, we obtain:

𝜎𝑖𝑗,𝑗 = (𝜆 + 𝜇)𝑢𝑘,𝑘𝑖 + 𝜇𝑢𝑖,𝑗𝑗

Its substitution in Eq. (a) yields the required Equation as:

(𝜆 + 𝜇)𝑢𝑘,𝑘𝑖 + 𝜇𝑢𝑖,𝑗𝑗 + 𝐹𝑖 = 0

𝜕𝑥
Example 2.9: Determine 𝜕𝑥 𝑖 for Cartesian coordinate system.
𝑗

Solution: Since 𝑥𝑖 , 𝑥𝑗 and 𝑥𝑘 are normal to each other, differentiation of 𝑥𝑖 with 𝑥𝑗 is zero but
differentiation of 𝑥𝑖 with 𝑥𝑖 is one. Thus,

𝜕𝑥𝑖
= 𝛿𝑖𝑗
𝜕𝑥𝑗

2.7 Matrix Multiplication using Summation Convention

Matrices are the language of vectors and second order tensors; in fact, some like to call them
as homes of tensors. The Einstein summation convention makes the matrix multiplication
very compact and convenient. We no longer need to write matrices in full form. It is worth
mentioning here that all matrices are not tensors, but tensors are expressed in terms of
matrices.

The basics of matrix multiplication is that the first row of first matrix is multiplied to the first
column of the second matrix, term by term, to obtain the first element value of the first row
of the resulting matrix. Similarly, other elements of the resulting matrix are obtained. Let us
consider the matrix 𝐶 which is obtained by multiplying matrix 𝐴 (2x3) with matrix 𝐵 (3x2) as:

𝐵11 𝐵12
𝐴11 𝐴12 𝐴13
𝐶 = 𝐴𝐵 = [ ] [𝐵21 𝐵22 ]
𝐴21 𝐴22 𝐴23
𝐵31 𝐵32

𝐴11 𝐵11 + 𝐴12 𝐵21 + 𝐴13 𝐵31 𝐴11 𝐵12 + 𝐴12 𝐵22 + 𝐴13 𝐵32
=[ ]
𝐴21 𝐵11 + 𝐴22 𝐵21 + 𝐴23 𝐵31 𝐴21 𝐵12 + 𝐴22 𝐵22 + 𝐴23 𝐵32

The requirement of the matrix multiplication is that the number of columns in the first matrix
should be same as the number of rows of the second matrix. When we focus on any element
of the resulting matrix, we find that in a term the second index of matrix 𝐴 is same as first
index of the matrix 𝐵 . Thus, in indicial form the matrix is written as:

𝐶𝑖𝑗 = 𝐴𝑖𝑘 𝐵𝑘𝑗

The dummy index 𝑘 must be the last index of the first matrix and the first index of the second
matrix for a valid matrix multiplication. An expression 𝐴𝐵𝒗 , where 𝐴 and 𝐵 are second order
tensors and v is a vector, can be written as:

𝐴𝐵𝒗 = 𝐴𝑖𝑘 𝐵𝑘𝑙 𝑣𝑙

A component form is commutative; that is, we can write:

𝐴𝑖𝑘 𝐵𝑘𝑙 𝑣𝑙 = 𝐵𝑘𝑙 𝐴𝑖𝑘 𝑣𝑙

However, whenever we multiply, we should rearrange the variables to meet the required
condition that the last index of first variable should be same as the first index of the second
variable. But in nonindicial form, the tensors are not commutative; that is:

𝐴𝐵 ≠ 𝐵𝐴

Consider an example where a multiplication is given as 𝐴𝑖𝑘 𝐵𝑗𝑘 . Clearly the dummy index k is
not in not located as the immediate neighbour and therefore we should manipulate it.

Transpose of a matrix comes out handy in such a situation. The transpose of a matrix is
obtained by converting rows to columns. If 𝐴𝑖𝑗 is expressed as:

𝐴11 𝐴12 𝐴13

𝐴𝑖𝑗 = [𝐴21 𝐴22 𝐴23 ]
𝐴31 𝐴32 𝐴33
its transpose is:

𝐴11 𝐴21 𝐴31

𝐴𝑇𝑖𝑗 = [𝐴12 𝐴22 𝐴32 ]
𝐴13 𝐴23 𝐴33

In compact form, the transpose is obtained just by interchanging the indices; that is:

𝐴𝑇𝑖𝑗 = 𝐴𝑗𝑖

Example 2.10: Express the term 𝐴𝑖𝑘 𝐵𝑗𝑘 in matrix multiplication form and write the matrices
in long form.

Solution: We take transpose of second matrix to bring its dummy index next to the dummy
𝑇
index of the first matrix, resulting into 𝐴𝑖𝑘 𝐵𝑘𝑗 . The long matrix form is:

𝐴11 𝐴12 𝐴13 𝐵11 𝐵21 𝐵31

𝑇 𝐵32 ] = 𝐴𝐵𝑇
𝐴𝑖𝑘 𝐵𝑗𝑘 = 𝐴𝑖𝑘 𝐵𝑘𝑗 = [𝐴21 𝐴22 𝐴23 ] [𝐵12 𝐵22
𝐴31 𝐴32 𝐴33 𝐵13 𝐵23 𝐵33

Comments/Interpretation: When we convert indicial form to nonindicial form, the indicial

expression should be arranged in a valid multiplication form.

Example 2.11: Determine 𝐴𝑖𝑗 𝐵𝑗𝑖 where A and B are tensors of second order.

Solution: There are two dummy indices and no free index and therefore it is a scalar. Let 𝐶 =
𝐴𝐵 which is written in indicial form as:

𝐶𝑖𝑘 = 𝐴𝑖𝑗 𝐵𝑗𝑘

Then 𝐶𝑖𝑖 = 𝐶11 + 𝐶22 + 𝐶33 . This is the sum of the diagonal terms of the matrix C. We can
multiply matrices A and B fully and then take the sum of the diagonal terms. Alternatively, we
just multiply only those rows and columns which yield the diagonal terms which is known as
Trace (AB). Thus,

𝐴11 𝐴12 𝐴13 𝐵11 𝐵12 𝐵13

𝐶 = Tr 𝐴𝐵 =Tr[𝐴21
( ) 𝐴22 𝐴23 ] [𝐵21 𝐵22 𝐵23 ]
𝐴31 𝐴32 𝐴33 𝐵31 𝐵32 𝐵33

= (𝐴11 𝐵11 + 𝐴12 𝐵21 + 𝐴13 𝐵31 ) + (𝐴21 𝐵12 + 𝐴22 𝐵22 + 𝐴23 𝐵33 )

+(𝐴31 𝐵13 + 𝐴32 𝐵23 + 𝐴33 𝐵31 )

Let us discuss how to represent the case of multiplication when a second order tensor A acts
on a vector v and the resulting vector is u. For example, consider the relation 𝒖 = 𝐴𝒗 which
is expressed in indicial notation as:
 𝑢𝑖 = 𝐴𝑖𝑗 𝑣𝑗

In long form it becomes:

𝑢1 𝐴11 𝐴12 𝐴13 𝑣1

{ 2 } = [𝐴21 𝐴22 𝐴23 ] {𝑣2 }
𝑢
𝑢3 𝐴31 𝐴32 𝐴33 𝑣3
(3x1) (3x3) (3x1)

In the above relation we have used vectors as column matrix. But if we consider expression
𝑣𝑖 𝐵𝑖𝑗 𝑢𝑗 , it is a scalar because there are two dummy indices and no free index and it can be
expressed in long form as:

𝐵11 𝐵12 𝐵13 𝑢1

{𝑣1 𝑣2 𝑣3 } [𝐵21 𝐵22 𝐵23 ] {𝑢2 }
𝐵31 𝐵32 𝐵32 𝑢3
(1x3) (3x3) (3x1)

This expression is also a scalar because there is no free index. It is worth noting here that a
vector can be expressed through a row matrix or a column matrix to meet the requirements
of the multiplication. In fact, this expression in nonindicial form is 𝒗. 𝐵𝒖. Tensor B acting on
vector u yields a vector and then the resulting vector is multiplied with vector v through the
dot product. You must have been exposed to dot and cross products of two vectors earlier
but we will discuss them through indicial notations.

2.8 Vector Products through the Summation Convention

Product of two vectors u and v is of two kinds: (i) dot product yielding a scalar, and (ii) cross
product yielding a vector. Both kinds are used in solid mechanics.
2.8.1 Dot product: In your earlier background, the dot product of vectors u and v is defined
as (Fig. 2.3):
𝒖. 𝒗 = 𝑢𝑣𝑐𝑜𝑠𝜃
 Fig. 2.3 The angle 𝜃 between the two vectors

In this relation, 𝜃 is the angle between the two vectors, and u and v are magnitudes of the
two vectors. Invoking the definition, we obtain the relation of dot product between two unit
normal as (Fig. 2.1a):

𝒆𝒊 . 𝒆𝒋 = 𝛿𝑖𝑗

Consider the vectors in long form as:

𝒖 = 𝑢1 𝒆𝟏 + 𝑢2 𝒆𝟐 + 𝑢3 𝒆𝟑
𝒗 = 𝑣1 𝒆𝟏 + 𝑣2 𝒆𝟐 + 𝑣3 𝒆𝟑
The dot product of these two vectors then becomes:

𝒖. 𝒗 = 𝑢1 𝑣1 + 𝑢2 𝑣2 + 𝑢3 𝑣3
In matrix form, the relation can be expressed as:
𝑣1
𝒖. 𝒗 = {𝑢1 𝑢2 𝑢3 } {𝑣2 }
𝑣3
(1x3) (3x1)
Through the summation convention this relation is expressed as:
𝒖. 𝒗 = 𝑢𝑖 𝑣𝑖
A good example of dot product is the work done on a body by force f and the resulting
displacement u. The work done W is conveniently expressed as:
𝑊 = 𝒇. 𝒖 = 𝑓𝑘 𝑢𝑘

Example 2.12: Determine the work done on a guided block (Fig 2.4) which moves by distance
Δ𝒖 = 140𝒆𝟏 − 12𝟎 + 325𝒆𝟑 mm if the force applied on the block is 𝒇 = 305𝒆𝟏 + 104𝒆𝟐 +
280𝒆𝟑 N
 (a) (b)

Fig 2.4 (a). Force applied on a guided block , and (b) Figure of Example 2.13

Solution: The work done W is a dot product given by:

𝑊 = 𝒇. Δ𝒖 = (305𝒆𝟏 + 104𝒆𝟐 + 280𝒆𝟑 ).(140𝒆𝟏 − 12𝟎 + 325𝒆𝟑 ) N.mm

= [305 × 140 + 104 × (−120) + 280 × 325] × 10−3 =121.2 J

Example 2.13: Consider force f and an unit vector n (Fig. 2.4b) given by

𝒇 = 10𝒆𝟏 + 15𝒆𝟐 + 40𝒆𝟑 N

2 2 1
𝒏 = 3 𝒆𝟏 + 3 𝒆𝟐 + 3 𝒆𝟑

Determine: (i) component of force in the direction of the unit vector, and (ii) the angle
between the force and the unit vector.

Solution:

(i). The component 𝑓𝑛 of force f in direction n is given by:

2 2 1
𝑓𝑛 = 𝒇. 𝒏 = 𝑓𝑖 𝑛𝑖 = 10 × 3 + 15 × 3 + 40 × 3 = 30 N

(ii). Angle 𝜃 is evaluated as:

𝑓𝑛 30
𝜃 = cos −1 (𝑓×1
𝑖 𝑖
) = cos −1 (√102 ) = 46.86o
+152 +402
2.8.2 Cross Product: The cross product of two vectors u and v is a vector and it is
conventionally defined as (Fig. 2.5):

𝒖 × 𝒗 = (𝑢𝑣𝑠𝑖𝑛𝜃)𝒆𝒏

In this definition, 𝒆𝒏 , is a unit vector normal which is normal to both vector u and vector v as
shown the figure.

(a) (b)

Fig. 2.5 (a) Unit vector 𝒆𝒏 is normal to the plane formed by vectors u and v, and (b) circle to decide
the sign of permutation symbol

We would now like to represent the cross product in indicial form. To achieve it, we create
one more symbol known as permutation symbol. It is represented by three indies as 𝜖𝑖𝑗𝑘
where i, j, k can take value from 1 to 3. It was first formulated by Italian Mathematician Tullio
Levi-Civita !873-1941).

The permutation symbol is defined to have zero value if two indices are equal. For example:

𝜖112 = 0 ; 𝜖322 = 0 ; 𝜖331 = 0

If all three indices are distinctly different its value is either I or -1. To define when it is I or -1,
consider a circle with 1, 2, 3 written on it as shown in Fig 2.5b. If the distinct indices are in
clockwise direction its value is +1. When the distinct indices are in counter clockwise, the
value is -1. For example:

Clockwise direction: 𝜖123 = 1 ; 𝜖231 = 1 ; 𝜖321 = 1

Counter clockwise direction: 𝜖321 = −1 ; 𝜖213 = −1 ; 𝜖132 = −1

The cross product of vectors u and v then becomes:

𝒖 × 𝒗 = 𝜖𝑖𝑗𝑘 𝑢𝑖 𝑣𝑗 𝒆𝒌 … (2.29)

Let us focus on right-side expression. If 𝑖 = 1 and 𝑗 = 2, the expression becomes 𝜖12𝑘 𝑢1 𝑣2 𝒆𝒌

and k must be 3 to have non-zero value of the permutation symbol having the resulting non-
zero term is 𝑢1 𝑣2 𝒆𝟑 . Thus, the right-side of Eq.2.29 represents six following non-zero terms:
 𝜖123 𝑢1 𝑣2 𝒆𝟑 = 𝑢1 𝑣2 𝒆𝟑

𝜖213 𝑢1 𝑣2 𝒆𝟑 = −𝑢2 𝑣1 𝒆𝟑

𝜖231 𝑢2 𝑣3 𝒆𝟏 = 𝑢2 𝑣3 𝒆𝟏

𝜖321 𝑢3 𝑣2 𝒆𝟏 = 𝑢3 𝑣2 𝒆𝟏

𝜖312 𝑢3 𝑣1 𝒆𝟐 = 𝑢3 𝑣1 𝒆𝟐

𝜖132 𝑢1 𝑣3 𝒆𝟐 = 𝑢1 𝑣3 𝒆𝟐

And Eq. 2.29 is expressed in long form as:

𝒖 × 𝒗 = (𝑢2 𝑣3 − 𝑢3 𝑣2 )𝒆𝟏 + (𝑢3 𝑣1 − 𝑢1 𝑣3 )𝒆𝟏 + (𝑢1 𝑣2 − 𝑢2 𝑣1 )𝒆𝟏

Some people like to like to represent the cross product as:

𝒆𝟏 𝒆𝟐 𝒆𝟑
𝒖 × 𝒗 = |𝑢1 𝑢2 𝑢3 |
𝑣1 𝑣2 𝑣3

A common example of cross product is the determination of a moment about a point. The
moment M about point O is defined as (Fig. 2.6a):

𝑴=𝒓×𝒇

Fig. 2.6 (a) Moment due to force f about point O, and (b) angular momentum

due to velocity v

It is noted that point Q can be anywhere on the force line because;

𝒓 × 𝒇 = (𝑟𝑓𝑠𝑖𝑛𝜃)𝒆𝒏 = (𝑂𝑀)𝒆𝒏

where 𝒆𝒏 is the normal unit vector and ON is the normal distance which remains unchanged
when point Q is moved on the force line. Another example is shown on the angular
momentum 𝒉 in Fig. 2.6b which is expressed again with the help of cross product of distance
𝒓 and linear momentum 𝑚𝒗 as:
 𝒉 = 𝒓 × (𝑚𝒗)

Example 2.14: Determine the angular momentum about point H of a body of 5 kg mass and
having velocity 𝒗 = 9𝒆𝟏 + 4𝒆𝟐 + 7𝒆𝟑 m/s. Point H is located at 𝒓𝑯 = 80𝒆𝟏 + 110𝒆𝟐 +
135𝒆𝟑 mm and the chosen point on the velocity line is 𝒓𝑸 = 180𝒆𝟏 + 410𝒆𝟐 + 218𝒆𝟑 mm
as shown in Fig. 2.7.

Fig. 2.7 Angular momentum about point H

Solution: Distance r of point Q from point H is

𝒓 = (180 − 80)𝒆𝟏 + (410 − 110)𝒆𝟐 + (218−𝟏𝟑𝟓)𝒆𝟑

= 100𝒆𝟏 + 300𝒆𝟐 + 83𝒆𝟑

Angular momentum h about point H is:

𝑘𝑔.𝑚 2
𝒉 = 𝒓 × (𝑚𝒗) =(100𝒆𝟏 + 300𝒆𝟐 + 83𝒆𝟑 ) × 5(9𝒆𝟏 + 4𝒆𝟐 + 7𝒆𝟑 ) × 10−3 𝑠

𝑘𝑔.𝑚 2
= 13.41 𝑠

Reviewing the vector product, we realize that both dot and cross products are widely used in
solid mechanics. A vector deals with the variables effectively in 3D-world we live in. To use
force, distance, velocity, acceleration, moment, etc. we should be well conversant to play with
vector algebra.

One of my friends works extensively on the geometrical aspects of designing structures. He

routinely divides the surface of the member into triangles. Quite often he needs to find the
area of such triangles. Knowing the coordinates of the three corner points of a triangle, he
determines the vectoral distance of any two sides. He then takes cross product of these two
vectors and divides the results by 2 to obtain the area of the triangle.
 2.9 Manipulations with Indicial Expressions

We will discuss three cases to be more familiar with indicial expressions.

Case1: For the first case consider the two equations:

𝑎𝑖 = 𝐴𝑖𝑘 𝑏𝑘 … (2.30)

𝑏𝑖 = 𝐵𝑖𝑘 𝑐𝑘 … (2.31)

In these two equations, a, b, c are vectors, and A and B are second order tensors. We want to
combine these equations to get rid of vector b. In Eq. 2.30, i is the free index and k dummy
index . To combine the equations, we change the index i of vector b to k to have:

𝑏𝑘 = 𝐵𝑘𝑙 𝑐𝑙

Note that the symbol k of dummy index in the Eq.2.31 must be changed to some other symbol,
l in this case. Now we substitute it in Eq. 2.30 to obtain the required relation as:

𝑎𝑖 = 𝐴𝑖𝑘 𝐵𝑘𝑙 𝑐𝑙

This operation can be carried out more easily in nonindicial form as:

𝒂 = 𝐴𝒃

𝒃 = 𝐵𝒄

Substituting the second equation in the first equation, we obtain:

𝒂 = 𝐴𝐵𝒄

This equation can be written in long form as:

𝑎1 𝐴11 𝐴12 𝐴13 𝐵11 𝐵12 𝐵13 𝑐1

𝑎
{ 2 } = [𝐴21 𝐴22 𝐴23 ] [𝐵21 𝐵22 𝐵23 ] {𝑐2 }
𝑎3 𝐴31 𝐴32 𝐴33 𝐵31 𝐵32 𝐵33 𝑐3

Case 2: we will be considering two ways to deal with the equation:

𝑢𝑖 = 𝑣𝑗 𝐵𝑖𝑗

A term in the indicial form is commutative. We can express the above equation as:
𝑢𝑖 = 𝐵𝑖𝑗 𝑣𝑗
(3x1) (3x3) (3x1)

In full matrix form:

𝑢1 𝐵11 𝐵12 𝐵13 𝑣1 𝐵11 𝑣1 + 𝐵12 𝑣2 + 𝐵13 𝑣3
𝑢
{ 2 } = [𝐵21 𝐵22 𝑣
𝐵23 ] { 2 } = {𝐵21 𝑣1 + 𝐵22 𝑣2 + 𝐵23 𝑣3 } … (2. 32)
𝑢3 𝐵31 𝐵32 𝐵33 𝑣3 𝐵31 𝑣1 + 𝐵32 𝑣2 + 𝐵33 𝑣3
The matrix multiplication can be carried out alternatively as:

𝑢𝑖 = 𝑣𝑗 𝐵𝑗𝑖𝑇
(1x3) (1x3) (3x3)

In this form vectors u and v are expresses row matrix to meet the requirements of matrix
multiplication. In full form:
𝐵11 𝐵21 𝐵31
{𝑢1 𝑢2 𝑢3 } = {𝑣1 𝑣2 𝑣3 } [𝐵12 𝐵22 𝐵32 ]
𝐵13 𝐵23 𝐵33
= {𝐵11 𝑣1 + 𝐵12 𝑣2 + 𝐵13 𝑣3 𝐵21 𝑣1 + 𝐵22 𝑣2 + 𝐵23 𝑣3 𝐵31 𝑣1 + 𝐵32 𝑣2 + 𝐵33 𝑣3 }

This is same as Eq. 2.32.

Case 3: Consider
𝜎𝑖𝑗 𝑛𝑗 − 𝜆𝑛𝑖 = 0

We want to express it in factorial form by taking n as a factor. The second term is modified to
obtain nj by taking the help of Kronecker delta as 𝛿𝑖𝑗 𝑛𝑗 to have:

𝜎𝑖𝑗 𝑛𝑗 − 𝜆𝛿𝑖𝑗 𝑛𝑗 = 0

Factoring it, we obtain:

(𝜎𝑖𝑗 − 𝜆𝛿𝑖𝑗 )𝑛𝑗 = 0

In full matrix form:

𝜎11 − 𝜆 𝜎12 𝜎13 𝑛1 0
[ 𝜎21 𝜎22 − 𝜆 𝜎23 ] {𝑛2 } = {0}
𝜎31 𝜎32 𝜎33 − 𝜆 𝑛3 0
This is an eigenvalue problem. To make some sense out of this equation, we must have:
𝜎11 − 𝜆 𝜎12 𝜎13
| 𝜎21 𝜎22 − 𝜆 𝜎23 | = 0
𝜎31 𝜎32 𝜎33 − 𝜆
In Chapter 3 on stress tensor, we will show that the eigenvalues are principal stresses and
corresponding eigenvectors are principal directions. In fact, in engineering field, eigenvalue
problems are quite common. For example, vibration modes deal with eigenvalues
eigenvectors, buckling load and modes are obtained through eigenvalue formulation and
velocity of stress waves in solid are eigenvalues.
In nonindicial form,
𝜎𝒏 − 𝜆𝒏 = 0
or, (𝜎 − 𝜆𝐼 )𝒏 = 0

where I is the identity matrix:

 1 0 0
𝐼 = [0 1 0 ]
0 0 1
Again |𝜎 − 𝜆𝐼| = 0 provides eigenvalues.

2.10 Transpose of a second order tensor

As mentioned already briefly, the transpose of a matrix is obtained by changing rows to
columns; that is, the first row becomes the first column, the second row becomes second
column and so on. In component form transpose of 𝑆𝑖𝑗 is 𝑆𝑗𝑖 ; we just interchange the indices.
In nonindicial form, the transpose of S is written as 𝑆 𝑇 .
We often need to transpose the product of the two second order tensors. We will show:
(𝐵𝐶 )𝑇 = 𝐶 𝑇 𝐵𝑇

If 𝐷 = (𝐵𝐶 )𝑇 , we can write it in indicial form as

𝑇
𝐷𝑖𝑗 = (𝐵𝑖𝑘 𝐶𝑘𝑗 )

Interchanging the free indices of on right-side term, we obtain:

𝐷𝑖𝑗 = 𝐵𝑗𝑘 𝐶𝑘𝑖

We now like to have 𝐶𝑘𝑖 before 𝐵𝑗𝑘 to have first free index on right-side term as the free index
𝑖 to be consistent with free indices of left-side. Also, we manipulate the indices to have the
matrices in the form of matrix multiplication. Thus,
𝑇 𝑇
𝐷𝑖𝑗 = 𝐶𝑖𝑘 𝐵𝑘𝑗

yielding,
(𝐵𝐶 )𝑇 = 𝐶 𝑇 𝐵𝑇

2.11 Transformation of Vector and Tensorial Components

We first need to understand what is a physical quantity before we start discussing

transformation. We will then pick up a discussion of transformation of vector components. It
will be followed by taking up transformation of second order tensor and higher order tensors.

2.11.1 Physical quantity: We start with an analogy. Person A may measure his orange juice in
litres and person B in quarts. The quantity of juice is same but measuring systems are
different. And we do need some kind of measuring system to quantify the volume of the
orange juice. The amount of juice is equivalent to physical quantity and the measuring system
is equivalent to coordinate system in solid mechanics. And the relation between the two
measures is equivalent to transformation relations

The distance d between London and Paris is a vector as it accounts for the magnitude and the
direction. The axes 𝑥1 , 𝑥2 and 𝑥3 we choose to deal with this vector are manmade. The
directions of these axes are chosen as per our convenience. For example, we can choose 𝑥1 as
the line joining London to Barcelona in Spain and other two axes, x2 and x3, can be also
chosen conveniently. Alternatively, one can choose the line joining London to Venice in Italy
to be 𝑥1′ , and then choosing 𝑥2′ and 𝑥3′ conveniently. In the London-Barcelona coordinate
system, the components of the distance vector are 𝑑1 , 𝑑2 and 𝑑3 while in the London to
Venice coordinate system, the components are 𝑑1′ , 𝑑2′ and 𝑑3′ . The distance between London
and Paris does not depend on the coordinate axes chosen and we can express the following:

𝒅 = {𝑑1 𝑑2 𝑑3 } = {𝑑1′ 𝑑2′ 𝑑3′ }

But 𝑑1 ≠ 𝑑1′ ; 𝑑2 ≠ 𝑑2′ , and 𝑑3 ≠ 𝑑3′ . It is worth noting that 𝑑1 , 𝑑2 and 𝑑3 collectively define
distance 𝒅. Similarly, 𝑑1′ , 𝑑2′ and 𝑑3′ collectively define the same distance 𝒅. Thus, distance 𝒅
is a physical quantity, independent of coordinate system. Similarly, vectors like velocity,
acceleration, force, angular velocity, moment, momentum, angular momentum, etc. are
physical quantities. In later chapters, we will study that second order tensors relate two
vectors and therefore the second order tensors are physical quantities too. Thus, stress and
strain which are second order tensors, are physical quantities. Further, fourth order stiffness
tensor 𝐶 in the relation 𝜎 = 𝐶𝜖 is also a physical quantity.

Components of one coordinate system can be related to those of another coordinate system.
For example, 𝑑1′ can be obtained from 𝑑1 , 𝑑2 , 𝑑3 and the angles between the two-coordinate
axes. Obtaining 𝑑1′ , 𝑑1′ and 𝑑1′ in terms of 𝑑1 , 𝑑2 and 𝑑3 is called transformation of components.
We need to do transformation quite often in sold mechanics.

2.11.2 Transformation of vector components

The transformation of vector and tensor components is important and we will develop its
analysis gradually, starting from 2D analysis. Figure 2. 8 shows that coordinate system, 𝑥1′ -
𝑥2,′ , which makes an angle 𝜃 with coordinate system, 𝑥1 -𝑥2 . The components of vector p are
p1 and p2 in 𝑥1 -𝑥2 system while they are p1’ and p2’ as shown. We will determine components
p1’ and p2’ in terms of p1, p2 and angle 𝜃. According to the theorem of projections, the sum of
projections of p1 and p2 on x1’-axis is equal to p1’. Thus,

𝑝1′ = 𝑂𝐸 + 𝑂𝐹 = 𝑝1 𝑐𝑜𝑠𝜃 + 𝑝2 𝑠𝑖𝑛𝜃

 Fig. 2.8 Transformation of vector components

𝑝1′ = 𝑂𝐸 + 𝑂𝐹 = 𝑝1 𝑐𝑜𝑠𝜃 + 𝑝2 𝑠𝑖𝑛𝜃

Similarly,

𝑝2′ = −𝑂𝐺 + 𝑂𝐻 = −𝑝1 𝑠𝑖𝑛𝜃 + 𝑝2 𝑐𝑜𝑠𝜃

These relations can be written in matrix form as:

𝑝′ 𝑐𝑜𝑠𝜃 𝑠𝑖𝑛𝜃 𝑝1
{ 1′ } = [ ]{ }
𝑝2 −𝑠𝑖𝑛𝜃 𝑐𝑜𝑠𝜃 𝑝2

In non-indicial form:

𝒑′ = 𝐴𝒑 … (2.33)

In this expression the matrix A is:

𝑐𝑜𝑠𝜃 𝑠𝑖𝑛𝜃
𝐴=[ ]
−𝑠𝑖𝑛𝜃 𝑐𝑜𝑠𝜃

Matrix A is called rotation matrix. We should understand and remember that vector p’ is
having its components p1’ and p2’ while vector p is associated with components p1 and p2. In
reality, p’ and p represent the same physical quantity. Also, matrix A is not a physical quantity;
its components just represent direction cosines of angles between the axes of the two-
coordinate system.
 Fig. 2 .9 Two sets of axes in the general case of 3D

We live in 3D-world and we solve many problems dealing with 3D structure members and,
therefore, we should develop the transformational relation for the general case of 3D-
structures. Figure 2.9 shows the two sets of the coordinate axes. We keep track of the angles
by the direction cosines as shown in the figure and Table 2.1.

Table 2.1: Direction cosines of angles between two sets of axes

𝑥1 𝑥2 𝑥3
𝑥1′ 𝑎11 𝑎12 𝑎13
𝑥2′ 𝑎21 𝑎22 𝑎23
𝑥2′ 𝑎31 𝑎32 𝑎33

The first row of the table shows the cosine of angle which axis x1’ makes with axes 𝑥1 , 𝑥2 and
𝑥3 . For example, a11 corresponds to cosine of the angle between x1 and x1’, a12 between x1’
and x2, and so. Similarly, the components of second row of the table shows cosine of angles
between x2’ and 𝑥1 , 𝑥2 and 𝑥3 . The table is written in indicial form as:

𝑥𝑖′ = 𝑎𝑖𝑗 𝑥𝑗 … (2.34)

where A is rotational matrix defined in Table 2.1. In the nonindicial form,

𝑥 ′ = 𝐴𝑥

I want to emphasize here again that matrix A is not a physical quantity although it is a 3x3
matrix. It depends only on three angles between the axes of the two coordinate systems
which are manmade. We use matrices to document tensors but all matrices are not tensors.

We can reverse our analysis by expressing 𝑥1 , 𝑥2 and 𝑥3 in terms of 𝑥1′ , 𝑥2′ and 𝑥3′ . The
directional cosines are shown in Table 2.2.
 Fig 2.10 Direction cosine of angles between x1’ and x1, x2, and x3.

Table 2.2 Direction cosines of axes 𝑥1 , 𝑥2 and 𝑥3 in terms of 𝑥1′ , 𝑥2′ and 𝑥3′

𝑥2′ 𝑥2′ 𝑥3
𝑥1 𝑎11 𝑎21 𝑎31
𝑥1 𝑎12 𝑎22 𝑎32
𝑥1 𝑎13 𝑎23 𝑎33

From the table, we obtain

𝑥𝑗 = 𝑎𝑘𝑗 𝑥𝑘′ … (2.35)

Substituting it in Eq. (2.34), we obtain:

𝑥𝑖′ = 𝑎𝑖𝑗 𝑎𝑘𝑗 𝑥𝑘′

When k is same as i, 𝑎𝑖𝑗 𝑎𝑘𝑗 = 1. For 𝑖 ≠ 𝑘 , 𝑎𝑖𝑗 𝑎𝑘𝑗 should be zero and we therefore can infer:

𝑎𝑖𝑗 𝑎𝑘𝑗 = 𝛿𝑖𝑘

𝑇
or, 𝑎𝑖𝑗 𝑎𝑗𝑘 = 𝛿𝑖𝑘

This is expressed in nonindicial form as:

𝐴𝐴𝑇 = 𝐼

or, 𝐴−1 = 𝐴𝑇
It is a simple task to take inverse of the rotational matrix A. The inverse is obtained just by
taking its transpose. Such a matrix is known as orthogonal matrix.

In the general case of 3D, the transformation relations for vector v are:

𝑝𝑖′ = 𝑎𝑖𝑗 𝑝𝑗 …. (2. 36)

𝑝𝑖 = 𝑎𝑗𝑖 𝑝𝑗′ = 𝑎𝑖𝑗

𝑇 ′
𝑝𝑗 … (2.37)

In nonindicial form:

𝒑′ = 𝐴𝒑 … (2.38)

𝒑 = 𝐴𝑇 𝒑′ … (2.39)

Example 2.15: If velocity v of a body is known as 𝒗 = 8𝒆𝟏 + 4𝒆𝟐 − 3𝒆𝟑 , determine the
transformed components of the new set of axes shown in Fig. 2.11.

Fig. 2.11
Solution: The direction cosine matrix is:
𝑥1 𝑥2 𝑥3
𝑥1′ 1 0 0
𝑥2′ 0 𝑐𝑜𝑠𝜃 𝑠𝑖𝑛𝜃
𝑥2′ 0 -sin 𝜃 cos 𝜃

Then the rotational matrix is:

1 0 0
𝐴 = [0 𝑐𝑜𝑠𝜃 𝑠𝑖𝑛𝜃 ]
0 −𝑠𝑖𝑛𝜃 𝑐𝑜𝑠𝜃

The transformed velocity components are:

𝑣1′ 1 0 0 8 8
′
{𝑣1 } = [0 𝑐𝑜𝑠𝜃 𝑠𝑖𝑛𝜃 ] { 4 } = { 4 cos𝜃 −3𝑠𝑖𝑛𝜃 }
𝑣1′ 0 −𝑠𝑖𝑛𝜃 𝑐𝑜𝑠𝜃 −3 −4𝑠𝑖𝑛𝜃 − 3𝑐𝑜𝑠𝜃

2.11.3 Transformation of second order components: A second order tensor is associated

with two directions. It is therefore expected that each direction should be accounted for the
transformation. Therefore, transformation of a second order tensor should involve two
rotational matrices A. I will first present the results and then we will go ahead with the proof.
If S is a second order tensor and the transformed tensor is S’ , they are related as:

𝑆 ′ = 𝐴𝑆𝐴𝑇

In indicial form:

𝑆𝑖𝑗′ = 𝑎𝑖𝑘 𝑆𝑘𝑙 𝑎𝑙𝑗

𝑇
= 𝑎𝑖𝑘 𝑆𝑘𝑙 𝑎𝑗𝑙 = 𝑎𝑖𝑘 𝑎𝑗𝑙 𝑆𝑘𝑙

To prove it, consider two vectors p and q related with a second order tensor S as:

𝑝𝑖 = 𝑆𝑖𝑙 𝑞𝑙 … (2.40)

𝑝𝒊′ = 𝑆𝑖𝑗′ 𝑞𝑗′ … (2.41)

We will now seek the relation between S and S’. Vector p’ is related to p as (Eq. 2.36):

𝑝𝑖′ = 𝑎𝑖𝑘 𝑝𝑘

Substituting Eq. 2.40 in this equation by first replacing free index i with k, we obtain:

𝑝𝑖′ = 𝑎𝑖𝑘 𝑆𝑘𝑙 𝑞𝑙

Transforming q to q’ using Eq, 2.37, we obtain

𝑝𝑖′ = 𝑎𝑖𝑘 𝑆𝑘𝑙 𝑎𝑗𝑙 𝑞𝑗′

This equation is compared with Eq. 2.41 to obtain:

𝑆𝑖𝑗′ = 𝑎𝑖𝑘 𝑎𝑗𝑙 𝑆𝑘𝑙

𝑇
= 𝑎𝑖𝑘 𝑆𝑘𝑙 𝑎𝑙𝑗

Its non-indicial form is:

𝑆 ′ = 𝐴𝑆𝐴𝑇

Example 2.16: Transform the components of a second order tensor σ if the axes are rotated
in x1-x2 plane as shown in Fig 2.12. The tensor is:

88 36 32
𝜎 = [36 49 29]
32 29 63

Figure 2.12 Two sets of axes

Table 2.3 Direction cosines between axes

𝑥1 𝑥2 𝑥3
′ 𝑜
𝑥1 𝑐𝑜𝑠22 𝑐𝑜𝑠(90 − 22 ) 𝑐𝑜𝑠90𝑜
𝑜 𝑜
′ 𝑜 𝑜
𝑥1 𝑐𝑜𝑠(90 − 22 ) 𝑐𝑜𝑠22𝑜 𝑐𝑜𝑠90𝑜
𝑥1′ 𝑐𝑜𝑠90𝑜 𝑐𝑜𝑠90𝑜 𝑐𝑜𝑠0𝑜

The orthogonal rotational matrix becomes:

0.927 0.375 0
𝐴 = [−0.375 0.927 0]
0 0 1

The transformed tensor σ’ is determined as:

0.927 0.375 0 88 36 32 0.927 −0.375 0

𝜎 ′ = 𝐴𝜎𝐴𝑇 = [−0.375 0.927 0] [36 49 29] [0.375 0.927 0]
0 0 1 32 29 63 0 0 1
 0.927 0.375 0 95.08 0.37 32
= [−0.375 0.927 0] [51.75 31.92 29]
0 0 1 40.54 14.88 63
107.5 12.31 40.54
= [12.31 0.927 14.88]
40.54 14.88 63.00

Comments/Interpretation: 𝑆 ′ = 𝐴𝑆𝐴𝑇 is a very powerful relation and quite suitable for a 3D

analysis. We will be transforming stress and strain components later in the book. Also, note
that a symmetric matrix is transformed to another symmetric matrix. This can be proved for
a general case; that is:

(𝑆 ′ )𝑇 = (𝐴𝑆𝐴𝑇 )𝑇 = 𝐴𝑆𝐴𝑇 because (𝐴𝐵𝐶 )𝑇 = 𝐶 𝑇 𝐵𝑇 𝐴𝑇 (2.42)

2.11.4 Transformation of higher order tensors: We will not carry out the rigorous derivation
but generalize the results obtained so far. For the vector, which is a first order tensor, and
second order tensors the transformation relations are:

𝑣𝑖′ = 𝑎𝑖𝑗 𝑣𝑗

𝑆𝑖𝑗′ = 𝑎𝑖𝑘 𝑎𝑗𝑙 𝑆𝑘𝑙

It is worth noting that for the first order tensor we need only one rotational matrix aij and
need two for a second order tensor. For a third order tensor, we need three orthogonal
matrices. If d is the third order tensor, the transformation relation is:
′
𝑑𝑖𝑗𝑘 = 𝑎𝑖𝑚 𝑎𝑗𝑛 𝑎𝑘𝑝 𝑑𝑚𝑛𝑝 … (2.43)

A relation for a tensor of order 3 or higher is not convenient to write in terms of matrix
multiplication as matrices are expresses only up to two indices only.

A common example of third order tensor is piezoelectric single crystal. When we apply stress
σ on it, a voltage expressed in terms of vector v is developed in the single crystal. Thus:

𝑣𝑖 = 𝑑𝑖𝑗𝑘 𝜎𝑗𝑘

During my PhD experimentation, once I used a piezoelectric single crystal to measure stress
in an impact experiment and I did use transformation relations.

Example 2.17: Consider a two-dimensional case of a third order tenor dijk; the indices vary
′
from 1 to 2 only. Determine 𝑑112 .

Solution: Eq. 2.43 for this case is:

′
𝑑112 = 𝑎1𝑘 𝑎1𝑙 𝑎2𝑚 𝑑𝑘𝑙𝑚

For 𝑘 = 1, we vary l from 1 to 2 and m also from 1 to 2 to have an expression for A1 as;
 𝐴1 = 𝑎11 𝑎11 (𝑎21 𝑑111 + 𝑎22 𝑑112 ) + 𝑎11 𝑎12 (𝑎21 𝑑121 + 𝑎22 𝑑122 )

For 𝑘 = 2, we again vary l from 1 to 2 and m also from 1 to 2 to have an expression A2 as:

𝐴2 = 𝑎12 𝑎11 (𝑎21 𝑑211 + 𝑎22 𝑑212 ) + 𝑎12 𝑎12 (𝑎21 𝑑221 + 𝑎22 𝑑222 )

Thus,
′
𝑑112 = 𝐴1 + 𝐴2

The transformation relation for a fourth order tensor Cijkl is:

′
𝐶𝑖𝑗𝑘𝑙 = 𝑎𝑖𝑚 𝑎𝑗𝑛 𝑎𝑘𝑝 𝑎𝑙𝑞 𝐶𝑚𝑛𝑝𝑞 … (2.44)

An example for fourth order tensor is the relation between the second order stress tensor σ
and the second order strain tensor 𝜖. They are related by the relation:

𝜎𝑖𝑗 = 𝐶𝑖𝑗𝑘𝑙 𝜖𝑘𝑙 … (2.45)

My postgraduate students routinely use 𝑆 ′ = 𝐴𝑆𝐴𝑇 or Eq. 2.42 while working with fiber
composite materials. We will discuss it in more detail in Ch. 7 on constitutive relations.

2.12 Differentiation

When I was a young boy in 1950s, a carpenter would come to my home to make a furniture
piece and he had a bag with some simple hand tools. Recently, I called a carpenter to make
wardrobes in my newly purchased flat, his tool box was quite sophisticated with an electrically
operated rotary cutter, a drill machine and a sander. Something similar happened to
mathematical tools in last 60 years or so. Scalar variables are upgraded to vectors and tensors.
This required upgradation of mathematical tools. We have already considered the indicial
equations with dummy summation indices, Kronecker delta, Permutation symbol and wide
spread usage of matrices. At the same time, calculus has been upgraded to handle vectors
and tensors.

A differential operator del (𝛻) has been developed which is defined as:

𝜕 𝜕 𝜕
𝛻 = 𝒆𝟏 𝜕𝑥 + 𝒆𝟐 𝜕𝑥 + 𝒆𝟑 𝜕𝑥
1 2 3

This operator is a vector that consists of partial derivatives and unit vectors. If it operates on
a scalar 𝜙, it yields a vector which is known as gradient of 𝜙. When 𝛻 operates on a vector as
a dot product, it is called divergence of a vector and the resulting expression is a scalar. When
it operates on a vector as cross product, it called curl of a vector to yield a vector. We will now
discuss all three operations in detail.

2.12.1: Gradient of a scalar: The gradient of a scalar 𝜙 is 𝛻𝜙 is expressed as:

𝜕 𝜕 𝜕
𝛻𝜙 = (𝒆𝟏 𝜕𝑥 + 𝒆𝟐 𝜕𝑥 + 𝒆𝟑 𝜕𝑥 )𝜙
1 2 3
 𝜕𝜙 𝜕𝜙 𝜕𝜙
= 𝒆𝟏 𝜕𝑥 + 𝒆𝟐 𝜕𝑥 + 𝒆𝟑 𝜕𝑥
1 2 3

𝜕𝜙
= 𝒆𝒊 𝜕𝑥
𝑖

The gradient of a vector has to be a vector because 𝛻 is a vector. In fact, 𝛻𝜙 is a vector normal
to the surface 𝜙 = 0. To prove it, consider a line on the surface 𝜙 = 0 as:

𝒅𝒙 = 𝑑𝑥1 𝒆𝟏 + 𝑑𝑥2 𝒆𝟐 + 𝑑𝑥3 𝒆𝟑

Then we take its dot product with 𝛻𝜙 to obtain:

𝜕𝜙 𝜕𝜙 𝜕𝜙
∇𝜙. 𝒅𝒙 =(𝒆𝟏 𝜕𝑥 + 𝒆𝟐 𝜕𝑥 + 𝒆𝟑 𝜕𝑥 ).( 𝑑𝑥1 𝒆𝟏 + 𝑑𝑥2 𝒆𝟐 + 𝑑𝑥3 𝒆𝟑 )
1 2 3

𝜕𝜙 𝜕𝜙 𝜕𝜙
= 𝜕𝑥 𝑑𝑥1 + 𝜕𝑥 𝑑𝑥2 + 𝜕𝑥 𝑑𝑥3 = 𝑑𝜙
1 2 3

On a surface, when we move from a point to a neighbouring point, 𝑑𝜙 = 0. Thus, ∇𝜙 is

normal to the surface.

Example 2.18: Determine normal to a spherical surface 𝜙 = 𝑥12 + 𝑥22 + 𝑥32 − 1 = 0 at point
1 1 1
( , , ). Also determine unit normal.
√3 √3 √3

Solution: Vector normal to the surface is:

𝜕𝜙 𝜕𝜙 𝜕𝜙
∇𝜙 = (𝒆𝟏 𝜕𝑥 + 𝒆𝟐 𝜕𝑥 + 𝒆𝟑 𝜕𝑥 )
1 2 3

= 2𝑥1 𝒆𝟏 + 2𝑥1 𝒆𝟏 + 2𝑥1 𝒆𝟏

The unit normal is then found out as:

∇𝜙 𝒆𝟏 𝒆𝟐 𝒆𝟑
𝒏 = |∇𝜙| = + +
√3 √3 √3

2.12.2 Divergence of a vector: Divergence of a vector v is expressed as:

𝜕 𝜕 𝜕
∇. 𝒗 = (𝒆𝟏 𝜕𝑥 + 𝒆𝟐 𝜕𝑥 + 𝒆𝟑 𝜕𝑥 ) . (𝑣1 𝒆𝟏 + 𝑣2 𝒆𝟐 + 𝑣3 𝒆𝟑 )
1 2 3

𝜕𝑣 𝜕𝑣 𝜕𝑣 𝜕𝑣
= 𝜕𝑥1 + 𝜕𝑥2 + 𝜕𝑥3 = 𝜕𝑥𝑖
1 2 3 𝑖

We now introduce one more convention which makes writing equations easier and still more
𝜕𝑀
compact. A partial differential is written as 𝑀,𝑖 and using this convention we write the
𝜕𝑥1
above equation as:
 ∇. 𝒗 = 𝑣𝑖,𝑖

2.12.3 Curl of a vector: The cross product of 𝛻 acting on a vector v yields a vector and is known
as curl of a vector. And is expressed as:

𝜕𝑣
∇ × 𝒗 = 𝜖𝑖𝑗𝑘 𝜕𝑥𝑖 𝒆𝒌 = 𝜖𝑖𝑗𝑘 𝑣𝑖,𝑗 𝒆𝒌
𝑗

In long form,

∇ × 𝒗 = (𝜖231 𝑣2,3 +𝜖321 𝑣3,2 )𝒆𝟏 + (𝜖312 𝑣3,1 +𝜖132 𝑣1,3 )𝒆𝟐 + (𝜖123 𝑣1,2 +𝜖213 𝑣2,1 )𝒆𝟑

= (𝑣2,3 −𝑣3,2 )𝒆𝟏 + (𝑣3,1 −𝑣1,3 )𝒆𝟐 + (𝑣1,2 −𝑣2,1 )𝒆𝟑

Example 2.19: Determine divergence and curl of vector 𝒗 = 𝑥1 𝑥22 𝒆𝟏 + 𝑥2 𝑥32 𝑒2 − 𝑥1 𝑥2 𝑥3 𝒆𝟑 .

Solution:

𝜕𝑣
Divergence: ∇. 𝒗 = 𝜕𝑥𝑖 = 𝑥22 + 𝑥32 − 𝑥1 𝑥2
𝑖

Curl: ∇ × 𝒗 = (2𝑥2 𝑥3 − 𝑥1 𝑥𝑥 )𝒆𝟏 + (−𝑥2 𝑥3 )𝒆𝟐 + (2𝑥1 𝑥2 )𝒆𝟑

2.12.4 Divergence of a second order symmetric tensor: It is defined in two ways but yield
same results if the second order tensor 𝜎 on which divergence is taken is symmetric. It is
defined as:

𝐷𝑖𝑣 𝜎 = 𝜎𝑖𝑗,𝑗 𝒆𝒊

In long form:

𝐷𝑖𝑣 𝜎 = (𝜎11,1 + 𝜎12,2 + 𝜎13,3 )𝒆𝟏 + (𝜎21,1 + 𝜎22,2 + 𝜎23,3 )𝒆𝟐 + (𝜎31,1 + 𝜎32,2 + 𝜎33,3 )𝒆𝟑

It is worth noting that divergence lowers the status of a tensor. When it acts on a vector, it
yields a scalar. When we take divergence of a second order tensor, the results in an expression
of a vector.

Some people like to write divergence of a tensor as ∇𝜎. It is worth noting that 𝐷𝑖𝑣 𝜎 is a
vector. In fact, a second order tensor is associated with two directions and when a dot product
is taken with a second order tensor, the resulting expression is associated with only one
direction. We will find in Ch. Xx that the equation of conservation of momentum is expressed
in terms of divergence of stress tensor.

Example 2.20: Determine divergence of the tensor 𝜎 given by:

𝑎𝑥22 𝑥1 𝑏𝑥1 𝑥2 𝑔𝑥32

𝜎 = [ 𝑏𝑥1 𝑥2 𝑐𝑥1 𝑥22 𝑘𝑥12 𝑥2 𝑥3 ]
𝑔𝑥32 𝑘𝑥12 𝑥2 𝑥3 𝑟𝑥12 𝑥3
Solution:

𝐷𝑖𝑣 𝜎 = 𝜎𝑖𝑗,𝑗 𝒆𝒊 = (𝑎𝑥22 + 𝑏𝑥1 + 2𝑔𝑥3 )𝒆𝟏 + (𝑏𝑥2 + 𝑐𝑥1 𝑥2 + 𝑘𝑥12 𝑥2 )𝒆𝟐 + (𝑘𝑥12 𝑥3 + 𝑟𝑥12 )𝒆𝟑

2.13 Gauss ( Divergence) Theorem

In solid mechanics, forces can be categorized into two types: (i) surface forces and (ii) body
forces. Surface forces are the forces applied at the surface of a structural member. Body
forces work on each and every point of the member and therefore it works on the entire
volume of the member.

Whenever we invoke a physical law, we account for all forces acting on a member. For
example, while applying the conservation of linear momentum, we sum all the surface and
body forces. The sum of surface forces is area integral because external forces act only on the
surface of the member. The sum of body forces is a volume integral. Thus, the balance
equation consists of both types of integral, surface and volume. To proceed further, we
convert all the surface integrals to volume integrals or vice versa. We therefore should have
a relation to achieve it. Gauss theorem facilitates this conversion.

Before the relation of the surface of the member of Gauss Theorem is stated, we will look into
the characteristics. In general, the orientation of a surface changes from a point to another
point. Consider an infinitesimal surface area dS as shown in Fig. 2.13. This small area of the
surface is characterized by its normal n at that point. We thus keep track of the normal when
we move from a point to another point of the surface. The normal plays a very important role
in solid mechanics. The vector n is expressed with its three component n1, n2 and n3.

Fig. 2.13 Normal at a point of the surface

Consider a surface integral with integrand 𝑀𝑛𝑖 as ∫𝑆 𝑀𝑛𝑖 𝑑𝑆 of a surface S which completely
encloses the volume V of the member. M can be a scalar, vector or tensor and ni represents
the normal dS. We will not prove the Gauss theorem. We will only state it as
 𝜕𝑀
∫𝑆 𝑀𝑛𝑖 𝑑𝑆 = ∫𝑉 𝜕𝑥 𝑑𝑉 = ∫𝑉 𝑀,𝑖 𝑑𝑉
𝑖

In nonindicial form:

∫𝑆 𝑀𝒏𝑑𝑆 = ∫𝑉 (𝐷𝑖𝑣 𝑀)𝑑𝑉

An analogy of the Gauss Theorem is a two storeyed building. The Gauss theorem is equivalent
to the stair case through which connects the ground floor to the first floor or the first floor to
the ground floor.

Example 2.21: Determine volume integral of the following: a) ∫𝑆 (𝑥12 + 3𝑥23 + 𝑥1 𝑥3 )𝑛𝑖 𝑑𝑆;

b) ∫𝑆[(𝑥1 𝑥22 𝑥3 )𝒆𝟏 + (𝟑𝑥1 𝑥2 𝑥3 )𝒆𝟐 + 𝑥1 𝑥22 𝑥3 )𝒆𝟑 ]𝑑𝑆; c) ∫𝑆 𝜎𝑖𝑗 𝑛𝑗 𝑑𝑆

Solution:

a) ∫𝑆(𝑥12 + 3𝑥23 + 𝑥1 𝑥3 )𝑛𝑖 𝑑𝑆 = ∫𝑉(𝑥12 + 3𝑥23 + 𝑥1 𝑥3 ),𝑖 𝑑𝑉

= ∫𝑉[(2𝑥1 + 𝑥3 )𝒆𝟏 + (6𝑥2 )𝒆𝟐 + 𝑥1 𝒆𝟑 ]𝑑𝑉

b) ∫𝑆[(𝑥1 𝑥22 𝑥3 )𝒆𝟏 + (𝟑𝑥1 𝑥2 𝑥3 )𝒆𝟐 + 𝑥1 𝑥22 𝑥3 )𝒆𝟑 ]. 𝒏𝑑𝑆 = ∫𝑉 𝐷𝑖𝑣[(𝑥1 𝑥22 𝑥3 )𝒆𝟏 +
(3𝑥1 𝑥22 𝑥3 )𝒆𝟐 + 𝑥1 𝑥22 𝑥3 )𝒆𝟑 ]𝑑𝑉 = ∫𝑉[(𝑥22 𝑥3 ) + (3𝑥1 𝑥3 ) + (2𝑥1 𝑥22 )]𝑑𝑉

c) ∫𝑆 𝜎𝑖𝑗 𝑛𝑗 𝑑𝑆 = ∫𝑉 𝜎𝑖𝑗,𝑗 𝑑𝑉 = ∫𝑉 𝐷𝑖𝑣 𝜎𝑖𝑗 𝑑𝑉

= ∫𝑉[(𝜎11,1 + 𝜎12,2 + 𝜎13,3 )𝒆𝟏 + (𝜎21,1 + 𝜎22,2 + 𝜎23,3 )𝒆𝟐 + (𝜎31,1 + 𝜎32,2 + 𝜎33,3 )𝒆𝟑

We often apply Gauss theorem to 2D-cases where an area is completely enclosed by its
perimeter s. Then the Gauss theorem is applicable for indices varying from 1 to 2 as:

𝜕𝑀
∫𝑠 𝑀𝑛𝑖 𝑑𝑠 = ∫𝐴 𝜕𝑥 𝑑𝐴 = ∫𝐴 𝑀,𝑖 𝑑𝐴
𝑖

Example 2.22: Convert an area integral to line integral given by:

𝜕(𝑥22+2𝑎𝑥1𝑥2 ) 𝜕(𝑥12+𝑐𝑥2 )
𝑁 = ∫𝐴[ + ]𝑑𝐴
𝜕𝑥2 𝜕𝑥1

Solution: The line integral is:

𝑁 = ∫𝑠[( 𝑥22 + 2𝑎𝑥1 𝑥2 )𝑛2 + (𝑥12 + 𝑐)𝑛1 ]𝑑𝑠

2.14 Summary

In solid mechanics, the problems are solved for 3D-cases with axes as x1, x2, x3 and associated
unit vectors e1, e2, e3. Indicial equations consisting of free and dummy indices are compact
way of writing many equations. If an equation does not have a free index, it is an equation
with all scalar terms. If there exists one free index, it represents a vectorial equation. Two free
indices in an equation means it is a relation of second order tensors. A dummy index stays
within a term. Kroneckar delta 𝛿𝑖𝑗 is equal to one for i=j and zero for 𝑖 ≠ 𝑗. It facilitates the
algebra carried out with indicial equations.

Dot and cross products are commonly used for the product of two vectors. Dot product yields
a scalar while cross product results into a vector. The permutation symbol 𝜖𝑖𝑗𝑘 is effective to
write cross product in indicial form.

Transformation of vector and tensor components is widely used in Solid mechanics. If A

represents the rotational matrix, transformed components of a vector v are obtained from
𝒗′ = 𝐴𝒗 and of a tensor S from the relation 𝑆 ′ = 𝐴𝑆𝐴𝑇

𝜕 𝜕 𝜕
Differentiation is performed by creating an operator del 𝛻 = (𝒆𝟏 𝜕𝑥 + 𝒆𝟐 𝜕𝑥 + 𝒆𝟑 𝜕𝑥 ).
1 2 3
When 𝛻 acts on a scalar, it gives a vector known as gradient. Dot product of 𝛻 with a vector
yield a scalar which is called divergence. Cross product with a vector yields a vector and is
known as curl. Operation of 𝛻 on a second order tensor S is expressed as Div S.

The Gauss theorem facilitates conversion of a surface integral of a closed body to volume
integral of the member fully enclosed by the surface. Its special case converts a line integral
to area integral.

2.15 Short Questions

1. How many equations does an indicial equation represent if it has two indicial free
indices?
2. Can we have more than one free index in a term? Give an example with a term having
2 free indices.
3. Can we change the symbol of the dummy index in a term?
4. In the product of two matrices, the last index of the first matrix should be same as the
first index of the second matrix. why so?
5. How do we find inverse of a rotational matrix?
6. If a term consists of three second order tensors what it is , a vector, a second order
tensor or higher order tensors?
7. A fourth order tensors relating two second order tensors is associated with four
directions. Why so?
8. In the transformation of the components of a second order tensor, we need to use
two rotational matrices. Why so?
9. Show that transformation of a second order symmetric tensor is a symmetric tensor.
10. Give examples of real-life cases which are expressed through dot product of two
vectors.
11. Give examples of real-life cases which are handled through cross product of two
vectors
12. Why is 𝛻𝜙 is a vector where 𝜙 is a scalar.
13. Express indicial form of curl of a vector.
14. Why is Gauss theorem useful in solid mechanics?

2.16 Problems

1. Write in long form:

(a) 𝐶𝑗𝑗 𝐷𝑘𝑘 ; (b) 𝑐𝑖 𝐷𝑘𝑘 ; (c) 𝑢𝑖 = 𝐴𝑖𝑗 𝑣𝑗

2. Write the following in terms of matrices. All matrices are 3x3. Do not multiply them.

(a) 𝐶𝑖𝑘 = 𝐴𝑖𝑗 𝐵𝑗𝑘 ; (b) 𝐷𝑖𝑘 = 𝐴𝑖𝑗 𝐵𝑘𝑗 ; (c) 𝑞 = 𝐶𝑚𝑚 + 𝐷𝑚𝑛 𝑎𝑚 𝑏𝑛 ;

(d) 𝐷𝑖𝑘 = 𝐴𝑖𝑘 𝐵𝑘𝑙 𝐶𝑗𝑙

3. Express in the long form using matrices: 𝒕 = 𝜎𝒏 where t and n are vectors and 𝜎
second order tensor.
4. Simplify the following; multiply matrices if needed:
1
(a) 𝐴𝑖𝑗 𝐴𝑗𝑖 ; (b) 𝐴𝑖𝑗 𝐴𝑖𝑗 ; (c) 𝑊 = 2 𝜎𝑖𝑗 𝜖𝑖𝑗 where 𝜎 and 𝜖 are symmetric 3x3 matrices.

5. If 𝑐𝑖 = 𝑑𝑖𝑗𝑘 𝐴𝑗𝑘 , determine 𝑐3

6. Transform stress components (Fig. 2.14) in 2D for:
73 42
𝜎=[ ] and axes are rotated by 33o
42 58

Fig 2.14
7. Write in index notation:

𝜕2∅ 𝜕 2∅ 𝜕2 ∅
+ 𝜕𝑥 + 𝜕𝑥 + 𝑐1 = 0
𝜕𝑥 2
1 1 𝜕𝑥2 1 𝜕𝑥2

𝜕 2∅ 𝜕2∅ 𝜕2∅
+ 𝜕𝑥 2 + 𝜕𝑥 + 𝑐2 = 0
𝜕𝑥1 𝜕𝑥2 2 2 𝜕𝑥3

𝜕2∅ 𝜕 2∅ 𝜕 2∅
+ + + 𝑐3 = 0
𝜕𝑥1 𝜕𝑥3 𝜕𝑥2 𝜕𝑥3 𝜕𝑥32
 8. Simplify:

(a) 𝑢𝑖 𝛿𝑖𝑗 ; (b) 𝛿𝑖𝑗 𝛿𝑗𝑖 ; (c) 𝜎𝑚𝑛 𝛿𝑚𝑖 𝛿𝑛𝑗 ; (d) 𝛿𝑖𝑗 𝛿𝑘𝑘 ; (e) (𝐴𝑖𝑗 − 𝐴𝑗𝑖 )𝛿𝑖𝑗

(f) 𝑢𝑖 𝑣𝑗 𝐵𝑘𝑙 𝛿𝑖𝑘 𝛿𝑗𝑙

9. Express all 9 equation with one indicial equation:

𝜕𝑢1 𝜕𝑢2 𝜕𝑢
𝐴11 = ; 𝐴22 = ; 𝐴33 = 𝜕𝑥3
𝜕𝑥1 𝜕𝑥2 3

𝜕𝑢 𝜕𝑢 𝜕𝑢 𝜕𝑢
𝐴12 = (𝜕𝑥1 + 𝜕𝑥2 ) ; 𝐴21 = (𝜕𝑥2 + 𝜕𝑥1 )
2 1 1 2

𝜕𝑢 𝜕𝑢 𝜕𝑢 𝜕𝑢
𝐴13 = (𝜕𝑥1 + 𝜕𝑥3 ) ; 𝐴31 = (𝜕𝑥3 + 𝜕𝑥1 )
3 1 1 3

𝜕𝑢2 𝜕𝑢3 𝜕𝑢3 𝜕𝑢2

𝐴23 = ( + ) ; 𝐴32 = ( + )
𝜕𝑥3 𝜕𝑥2 𝜕𝑥2 𝜕𝑥3

10. Show:

(a) 𝜖𝑖𝑗𝑘 𝑏𝑗 𝑏𝑘 = 0 ; (b) 𝑎 × 𝑏. 𝑐 = 𝜖𝑖𝑗𝑘 𝑎𝑖 𝑏𝑗 𝑐𝑘

11. Consider angles between the axes for transformation as

𝑥1 𝑥2 𝑥3
𝑥1′ 90𝑜 45𝑜 135
𝑥1′ 45𝑜 60𝑜 60𝑜
𝑥1′ 45𝑜 90𝑜 90𝑜

a) Determine transformation matrix A

b) Show 𝐴𝐴𝑇 = 𝐼
c) Transform components of stress tensor 𝜎:

20 −20 15
𝜎 = [−20 60 0]
15 0 40

12.Determine normal unit vector of curve 𝑥22 = 4𝑥1curve at point (1, -2)

13. Determine work done if force 𝒇 = 3𝒆𝟏 + 4𝒆𝟐 − 3𝒆𝟑 N moves a member by distance
𝑠 = 0.1𝒆𝟏 + 0.054𝒆𝟐 + 0.07𝒆𝟑 m
14. Determine angular momentum G if mass m=1.6 kg, velocity v=6-6 m/s and distance
from the point about which angular momentum is determined to a point on the v is
𝒓 = 0.3𝒆𝟏 − 0,5𝒆𝟐 + 0.7𝒆𝟑 m
15. Convert the surface to volume integral:
a) ∫𝑆 𝜎𝑖𝑗 𝑛𝑗 𝑑𝑆
b) ∫𝑆 𝜖𝑖𝑗𝑘 𝑥𝑗 𝜎𝑘𝑙 𝑛𝑙 𝑑𝑆 check