Está en la página 1de 26

Introduction to Continuum Mechanics

Lecture Notes
Jagan M. Padbidri
School of Engineering Sciences

Mahindra Ecole
Centrale

Fall 2016

Chapter 1
Essential mathematics
1.1

Vector algebra and indicial notation

Quantities that can be completely described by a single number are referred to as scalars.
These quantities are independent of any direction and physically, the number that denotes them refers to the magnitude of the quantity. Typical examples for scalars are
mass, temperature, density etc. To describe quantities that are associated with a specific direction along with a certain magnitude, vectors are used. Usually, one associates
vectors with the description of physical quantities such as force, velocity, acceleration etc.
Since our motivation is to describe physical phenomena using vectors, we will confine
ourselves to real valued vectors. The same holds true for scalars as well, i.e., we are
interested in real valued scalars. The set of all possible such scalars thus forms the real
value space R. We also define the set of all real valued vectors to form the space V. Both
these spaces are linear spaces and would need to satisfy certain conditions. Before listing
the conditions, we will follow that vectors will be designated using a bold face font or
an underscore and unit vectors will be identified with a carat () symbol. Real valued
vectors belonging to the linear vector space V will need to satisfy the following properties.
Additive commutativity:
For any arbitrarily chosen vectors u, v V,
u+v=v+u
Here, it is also implied that the resultant sum is a vector and is also a member of the
vector space V. Thus, the addition of vectors is an operation that takes two elements of
the vectors space V and maps the resultant, also, on to the space V, i.e., this operation
can be described as VV V. The represents a generic operation, not mathematical
multiplication.
Additive associativity:

For any arbitrarily chosen vectors u, v, w V,


(u + v) + w = u + (v + w)
This is also a V V V operation.
Zero element:
There exists an element 0 V such that
u+0=u
Negative element:
For every u V, there exists a v V such that
u + v = 0 = v = -u
Identity scalar multiplication:
For every u V,
1u=u
Note that since 1 R , the above operation involves a scalar operating on a vector to
result a vector, i.e., an R V V operation.
Distributivity with vector addition:
For all R and u, v V,
(u + v) = u + v
Distributivity with scalar addition:
For all , R and u V,
( + ) u = u + u
Associativity with scalar multiplication:
For all , R and u V,
() u = ( u)
Scalar or Dot product
The dot product, or the scalar product of two vectors u, v V is given by
u v = |u| |v| cos()
where |u| and |v| are the magnitudes of the vectors u and v and is the angle between
them. Physically, the dot product of u and v represents the magnitude of v multipled
with the component of u in the direction of v. The converse is true as well. For the
special case of = 0, cos() = 1, we have
u u = |u||u| = |u|2

= |u| = u u

(1.1)

The dot product is an operation on two vectors that results in a scalar, i.e., it is a
V V R operation. Extending the possible values assumed by ,
( = /2)
u v = 0 = u v
We note the dot product also satisfies the following conditions
uv=vu

Commutativity

u (v + w) = u v + u w

Distributivity

(u v) = u (v)

Associativity with scalars

Figure 1.1: (a)Dot product of two vectors and (b) Cross product of two vectors.
Vector or Cross product
The cross product or the vector product of two vectors, u and v results in another vector.
The magnitude of the cross product is given by |u v | = |u| |v| sin() where |u| and |v|
are the magnitudes of the vectors u and v and is the angle between them. The result
of the cross product is a vector w which is perpendicular to both u and v. Physically, the
cross product of u and v represents the area of the parallellogram formed by u and v.
|w| = u v = |u| |v| sin()
The orientation of the vector w is governed by the right hand rule. The cross product of
two vectors satisfies the following conditions
u v = v u

Anti-Commutativity

u (v + w) = u v + u w

Distributivity

(u v) = u (v)

Associativity with scalars

Scalar triple product


The triple product or box product of three vectors u, v and w is given by
(u v)w = (v w)u = (w u)v
If the three vectors u, v and w form a right handed triad, then the triple product gives
the volume of the parallellopiped formed by them.
Formation of a basis
The general rules for addition of vectors belonging to linear vector space can be extended
to an arbitrary set of vectors, i.e., the sum of an arbitrary set of vectors that belong to V
should result in a vector belonging to V. This can be written in a general form as follows.
Given a set of scalars (1 , 2 , . . . , n ) R and a set of vectors (u1 , u2 , . . . , un ) V,
3

Figure 1.2: Schematic for scalar triple product.


1 u1 + 2 u2 + + n un = v V
For the specific case of v = 0,
1 u1 + 2 u2 + + n un = 0
If for an arbitrary set of vectors (u1 , u2 , . . . , un ) V, we can find a set of scalars
(1 , 2 , . . . , n ) R, not all zero such that the above relation is satisfied, the set
(u1 , u2 , . . . , un ) is said to be linearly dependent.
On the other hand, if
1 u1 + 2 u2 + + n un = 0
implies that 1 = 2 = = n = 0, then the set (u1 , u2 , . . . , un ) is said to be linearly
independent.
Now, if we have a set of linearly independent vectors, (u1 , u2 , . . . , un ) and we chose a set
of scalars, (1 , 2 , . . . , n ), not all 0, then
1 u1 + 2 u2 + + n un 6= 0
Rather,
1 u1 + 2 u2 + + n un = v V
There are two important conclusions that can be drawn from this. First, since
(1 , 2 , . . . , n ) are chosen arbitrarily and v V, this means that for any v V, a
set of (1 , 2 , . . . , n ) can be obtained to satisfy the above relation. Second, for a given
v, the set (1 , 2 , . . . , n ) is unique since (u1 , u2 , . . . , un ) is linearly independent. This
means that any arbitrary v V can be desribed using the linearly independent set
(u1 , u2 , . . . , un ), i.e., (u1 , u2 , . . . , un ) acts as a basis for the description of elements in V.
Such a set of vectors are called the basis vectors for the space V.
The most commonly used set of basis vectors are the ones used to describe vectors in
the cartesean coordinate system. While a vector space is not limited by dimension, since
the cartesean system describes 3D Euclidean space (R3 ), there are three basis vectors in
this system. We commonly refer to them as the x, y and z axes. Since these are basis
vectors, any vector in (R3 ) can be represented as a linear combination of these three basis
4

vectors.

Figure 1.3: Example of a basis in R3 .


To simplify determining the set of scalars required to express an arbitrary vector in
terms of the basis vectors, we usually impose the condition that the basis vectors are of
unit magnitude, i.e., the vectors along the x, y and z directions are in fact x, y and z
Also, the vectors that form the basis are orthogonal to each other, i.e.,
or i, j and k.
i j = j k = i k = 0. Hence, this basis is called an Orthonormal basis. Now, any vector
v R3 space can be expressed as a linear combination of i, j and k as
v = vxi + vy j + vz k

(1.2)

where vx , vy and vz are the components of v along i, j and k and represent the weight
of the vector along said directions. It follows that
vx = v i; vy = v j; vz = v k

(1.3)

Figure 1.4: Example of representation of a vector v in a basis.


Note that it is not necessary that a vector space has a unique basis. It can have
infinitely many basis. However, the representation of any vector v R3 is unique for
that specific basis. We now revisit the vector operations previously addressed. Consider
arbitrary vectors u and v R3 . The scalar product of the vectors can be written as
(vxi + vy j + vz k)

u v = (uxi + uy j + uz k)
= uxi vxi + uxi vy j + uxi vz k
+ uy j vxi + uy j vy j + uy j vz k
+ uz k vxi + uz k vy j + uz k vz k
= u v = ux vx + uy vy + uz vz
5

(1.4)

Further, we can write the cross



i j


u v = ux uy

vx vy

product of u and v as

k

uz

vz

= u v = (uy vz uz vy )i (ux vz uz vx )j + (ux vy uy vx )k

(1.5)

Index notation
The above examples illustrate that even simple vector operations are rather cumbersome
in representation. We now introduce the notion of using indices to simplify representation
of vectors and avoid writing lengthy expressions. Since each vector is represented using
the basis vectors, we need to first address representing the basis vectors themselves in
with the set
the indicial notation. We replace the set of basis vectors (
x, y, z) or (i, j, k)
(
e1 , e2 , e3 ). This set can now be generically represented as ei where i = 1, 2, 3. Note that
the index i serves as a placeholder to assume values 1, 2 or 3 and has no relation to i.
We can express a vector as
v = vxi + vy j + vz k
= v1 e1 + v2 e2 + v3 e3
=

3
X

vi ei

i=1

We further simplify the notation by introducing a summation convention: A repeated


index automatically implies summation over all possible values for that index. Using this
convention, the last term of the above equation automatically implies summation over all
the possible values for i even if summation is not explicitly specified. Given that we are
concerned with 3D Euclidean space, i would take values of 1, 2 and 3 implying
v = vi ei = v1 e1 + v2 e2 + v3 e3

(1.6)

Note that the choice of using i as the index is arbitrary. Any other alphabet could be
used as an index, per convenience. We have already specified that the index i is not
related to i. As we have detailed previously, ei actually represents a set of three vectors
Similarly, ej represents the set of three vectors, albiet the exact
(
e1 , e2 , e3 ) or (i, j, k).
number that i and j assume can be arbitrary.
Now, consider the simple operation of finding the dot product of two unit vectors
ei and ej . We have already addressed the nature of the dot product when considering
vectors from an orthonormal triad. Just to summarize, ei and ej , each represent the
relevant unit vector depending on the value of the index and we are interested in finding
ei ej . Since ei and ej each represent three vectors, we have a total of nine possibilities.

They are
i = 1, j = 1; ei ej = e1 e1 = 1
i = 1, j = 2; ei ej = e1 e2 = 0
i = 1, j = 3; ei ej = e1 e3 = 0
i = 2, j = 1; ei ej = e2 e1 = 0
i = 2, j = 2; ei ej = e2 e2 = 1
i = 2, j = 3; ei ej = e2 e3 = 0
i = 3, j = 1; ei ej = e3 e1 = 0
i = 3, j = 2; ei ej = e3 e2 = 0
i = 3, j = 3; ei ej = e3 e3 = 1
The use of indicial notation allows us to account for all of these possibilities into the
compact form ei ej , of course with the knowledge that i and j could take values between
1 and 3. We can also make another, rather obvious observation from the above set of
equations that
ei ej = 1

if i = j

=0

otherwise

We simplify this by introducing a more convenient symbol called the Kronecker delta
given by
ei ej = ij = 1
=0

if i = j
otherwise

(1.7)

Note that ij has two indices i and j, each of which could take three possible values, i.e.,
ij can account for all the nine possibilities shown above. Of course, of these possibilities,
only three have non-zero values (i = j = 1; i = j = 2; i = j = 3). The summation
convention for the repeated indices that we have previously introduced, holds good here
as well i.e.,
ii = 11 + 22 + 33 = 3
We note that the notion of finding the dot product of basis vectors can be used directly
to find the dot product of any two vectors. Since any arbitrary vector can be expressed
in terms of a set of scalar components and the basis vectors, we can use the properties

of linear vector spaces as,


u v = ui ei vj ej
= ui vj ei ej
= ui vj ij
The above expression, ui vj ij implies that it is a summation over the repeated indices i
and j. Since both i and j can take values of 1, 2 or 3, the terms of the summation can
be written as
ui vj ij = u1 v1 11 + u1 v2 12 + u1 v3 13
+ u2 v1 21 + u2 v2 22 + u2 v3 23
+ u3 v1 31 + u3 v2 32 + u3 v3 33
This can be rearranged as
ui vj ij = [u1 v1 11 + u2 v2 22 + u3 v3 33 ]
+ u1 v2 12 + u1 v3 13
+ u2 v1 21 + u2 v3 23
+ u3 v1 31 + u3 v2 32
All the s outside the brackets reduce to 0
ui vj ij = [u1 v1 11 + u2 v2 22 + u3 v3 33 ] + 0
ui vj ij = u1 v1 + u2 v2 + u3 v3
= ui vj ij = ui vi
Comparing the respective terms, we get vj ij = vi
Alternatively, let us consider just the term vj ij . For any given value of the index
i, vj ij represents summation over three possible values of j, i.e., vj ij is the sum of
three terms with j taking on values 1, 2 and 3. In these three terms, there will be only
one instance when both i and j have the same value. For the remaining two terms,
i 6= j = ij = 0 and consequently, those two terms vanish! The only non-zero term in
the summation corresponds to when i = j = ij = 1 for that particular term. Thus,
the summation vj ij condenses to just one term, and i = j for that term and we can write
vj ij = vi

(1.8)

Thus, ij acts on specific components of v and effectively flips the index of that component by rendering the remaining terms of the summation to 0. The dot product of two
vectors can be written as
u v = ui vj ij
= u v = ui vi
8

(1.9)

Since the above expression has a repeated index, we can write out the summation over
that index and verify that the indicial notation does represent the dot product
u v = ui vi = u1 v1 + u2 v2 + u3 v3
Now, let us consider the cross product of two vectors.
u v = ui ei vj ej = ui vj ei ej
ui and vj are components of the vectors u and v i.e., they are scalars. The challenge lies
in describing the vector product ei ej . This is seemingly more complex than describing
the dot product of two basis vectors since the result of a dot product is a scalar quantity,
but we know that the result of a cross product is a vector, subject to the right hand rule.
Let us write down the possibilities for ei ej .
i = 1, j = 1; e1 e1 = 0
i = 1, j = 2; e1 e2 = e3
i = 1, j = 3; e1 e3 =
e2
i = 2, j = 1; e2 e1 =
e3
i = 2, j = 2; e2 e2 = 0
i = 2, j = 3; e2 e3 = e1
i = 3, j = 1; e3 e1 = e2
i = 3, j = 2; e3 e2 =
e1
i = 3, j = 3; e3 e3 = 0
We observe certain patterns in the result of the vector product of two basis vectors. When
i = j, the vector product is 0. For the instances when i 6= j and the cross product is
non-zero, the indices i and j yield a vector which corresponds to the third index, k 6= i, j,
i.e., e2 e3 = e1 , for example. Also, specific sequences of i and j produce a positive vector
in the right handed system and certain other sequences produce a negative vector. For
example, when i and j belong to the sequence 1, 2, 3, i.e., (i, j = 1, 2; i, j = 2, 3; i, j = 3, 1),
the result is a positive vector along ek , and when i and j are in the cyclically opposite
sequence (3, 2, 1), i.e., (i, j = 2, 1; i, j = 3, 2; i, j = 1, 3), the result is a negative vector
along ek . This can be summarized as
ei ej = ijk ek
where ijk is called the permutation symbol whose value is given by
ijk =1

if i, j, k belong to 1, 2, 3 sequence,

=1

if i, j, k belong to 3, 2, 1 sequence,

=0

otherwise.
9

(1.10)

We can now write the cross product of two vectors u and v as


w = u v = ui ei vj ej
= ui vj ei ej
= ui vj ijk ek
= w = u v = ijk ui vj ek

1.2

(1.11)

Transformation of vectors, Introduction to tensors

Transformation of vectors
We have mentioned in the previous section that vectors have both magnitude and
direction and have seen the representation of vectors in an orthonormal triad of basis
vectors. However, we have not specifically addressed the uniqueness of this basis. It is
rather obvious that the three dimensional real Euclidean space (R3 ) we live in does not
have a unique set of basis vectors i.e., one can define a set of basis vectors per convenience
and can also switch between the bases so defined. Note that we are referring to changing
the basis for a given vector, not changing or altering the vector itself. This is reflected as
a change in the representation of the vector, but not the vector itself. For example, if we
have a vector v and two different bases ei and ej , the above argument can be summarized
as
v = vi ei = v
j ej = vi 6= vj if ei 6= ej
This still gives no information on relating vi and vi or even ei and ei . We will address
this by formulating the transformation of a vector from one basis to another.
Consider a basis ei = (
e1 , e2 , e3 ) that has been rotated arbitrarily to obtain a new
basis ei = (e1 , e2 , e3 ), as shown in Fig. 1.2. Note that both the bases are right handed
orthonormal triads. Since the basis vectors are not perfectly alligned with each other,
each of the old basis vectors, ei will have a non-zero component along each of the new
basis vectors, ei . The magnitudes of these components can be obtained from the dot
products of the relevant basis vectors, i.e., we can express each of the new basis vectors
as if they were an arbitrary vector in the old basis.
e1 = (e1 e1 )
e1 + (e1 e2 )
e2 + (e1 e3 )
e3
= cos(e1 , e1 )
e1 + cos(e1 , e2 )
e2 + cos(e1 , e3 )
e3
= cos 11 e1 + cos 12 e2 + cos 13 e3
= e1 = a11 e1 + a12 e2 + a13 e3

10

Figure 1.5: Transformation of basis vectors.


It is emphasized here that one needs to be attentive to the conventions being used
to describe the angle between the new and the old basis. Here represents the angle
between one new basis vectors and one old basis vector. The first index for represents
the new basis, ei and the second index represents the old basis, ei , i.e., 23 represents
the angle between e2 and e3 . The other basis vectors in the ei basis follow a similar
mathematical pattern with respect to the vectors belonging to the ei basis. Thus, the
relations between the two bases can be expressed as


e1
e1
a11 a12 a13


e2 = a21 a22 a23 e2
e3
a31 a32 a33
e3
or in the more succint indicial notation as ei = aij ej . We denote the matrix shown above
as A, the transformation matrix. Similarly, we can express the old basis vectors in terms
of the new ones as
e1 = cos(e1 , e1 )e1 + cos(e2 , e1 )e2 + cos(e3 , e1 )e3
= cos 11 e1 + cos 21 e2 + cos 31 e3
e1 = a11 e1 + a21 e2 + a31 e3
= ei = aji ej
Now, we know that the new basis vectors, ei form an othonormal triad, i.e., ei ej = ij .
We can write
ei ej = aik ek ajl el
= aik ajl ek el
= aik ajl kl
= aik ajk
ij = aik ajk
ij = aik (akj )T
= AAT = I

11

This yields us the very important observation that the transpose of the transformation
matrix is its inverse, i.e., the transformation matrix is orthogonal.
Now, consider a vector that is represented in both the bases - v
in the ei basis and is

equivalent to v in the ei basis. Using the previously obtained property of the transformation matrix,
v
= vj ej v = vi ei = vi aji ej

By comparing the relevant coefficients,


vj = aji vi = v
= Av = vAT

The above equation prescribes the rules that need to be satisfied by entities to qualify
themselves to be vectors, i.e., vector quantities have to transform from one basis to another described by the relations above. We emphasize here that we are merely changing
the basis for a given vector, not altering the vector itself. From the above equations, we
see that the components of a vector change with a change in the basis in which it is being
represented. Then, what do we mean by the statement that the vector itself has not
changed? As we shall see, there are certain properties of vectors that are immune to the
transformation from one basis to another. In fact, the property of vectors to transform
from one basis to another through a linear mapping preserves the magnitude and the
direction of the vector.
The idea that a vector has not changed or varied, i.e., remained invariant, implies that
the representation and method of quantification of a vector have not changed. For example, consider the representation of the vector described above in both the bases, ei and ei .

The magnitude of the vector in the basis ei is given by |v| = v v = vi vi and in the

basis ei is |
v| = v
v
= vi vi . It is straighforward to prove that the transformation


of components between the two vectors, as derived above ensures that the magnitude is
the same. However, more fundamental is the expression for calculating the magnitude
of the vector. We see that the method/ expression to calculate the magnitude remains
the same irrespective of the basis being used i.e., the expression is invariant to the basis.
This is the fundamental argument towards the invariance of a vector with respect to the
basis. The reader is advised to repeat the process to explore the other quintessential
property of a vector, its direction. Of course, since the basis has been changed, so will
the orientation of the vector described in that basis. Thus, the invariance of the direction
of a vector needs to be verified using another arbitrarily defined vector.
We have discussed in detail the transformation of vectors from one basis to another.
This can be easily extended to describe rotation of vectors in the same basis. Instead of
keeping our vector fixed and rotating the basis vectors to a new position, we can keep
the basis fixed and define a new vector by rotating the existing vector about the basis
12

vectors. Determining the components of this new vector in the same basis is analogous
to the previous case where the vector was fixed and the basis was rotated. Thus, the
transformation matrix defined above also doubles up as a rotation matrix. Such an operation maps a vector from an initial position in the basis to a final position/ state.
Thus, the rotation matrix acts as a linear operator that maps one vector into another.
Introduction to Tensors
The idea of mapping one vector to another using linear operators is common place,
for eg., the scalar mass is used to map the force and acceleration vectors. However, the
spectrum of mapping one vector to another has been left incomplete i.e., we have considered the more convenient examples so far. Mapping the force and acceleration vectors
takes place through a physically measurable scalar and is very relavant physically, with
both vectors oriented along the same direction. We have discussed mapping one vector to
another using the rotation matrix with the directions of the two vectors being different,
but the rotation matrix itself does not correspond to any physically measurable quantity.
We now introduce quantities which map/ transform vectors and have physical significance
- Tensors.
Consider the following description of the angular momentum of a particle. Let the
particle have a mass m, position r = r1 e1 + r2 e2 and angular velocity = 3 e3 . We are
familiar that the angular momentum vector of the particle is given by h = I = mrv,
where v = r is the velocity of the particle. With a basic check, we can see that for r
in the (
e1 , e2 ) plane and along e3 , the velocity, v also lines in the (
e1 , e2 ) plane. The
angular momentum vector, h = mrv is oriented along e3 . Thus, the role of the moment
of inertia, I is analogous to mass m in F = ma i.e., it maps to h with both vectors along
the same direction. We note that there exist significant differences between both cases
since the force and acceleration vectors are defined with respect to the origin of the basis,
while angular momentum and angular velocity can be defined about an arbitrary point
or an axis that does not align with any individual basis vector. Nevertheless, angular
momentum and angular velocity are vectors which are mapped onto each other by the
moment of inertia.
Now, consider that the particle has the same angular velocity, = 3 e3 , but its
position vector is given by r = r1 e1 + r2 e2 + r3 e3 . We can deduce that v = r is still
parallel to the (
e1 , e2 ) plane. However, the angular momentum vector, h = mrv will not
e1 , e2 ) plane. So here, the moment of inertia maps
be along e3 since r does not lie in the (
to h which are along different directions. This deviates from our previous observation
of I being analogous to m and one can deduce that I is not a mere scalar. Let us analyse

13

Figure 1.6: Schematic of a rotating mass and associated vectors.


the expression for angular momentum in detail. We know,
h = mr v
= mr ( r)
= m(r r) m(r )r
Written in the indicial notation, we get
h = hi ei = (mrj rj )i ei (mk rk )ri ei
= (mr1 2 + mr2 2 + mr3 2 )1 e1 (m1 r1 + m2 r2 + m3 r3 )r1 e1
+ (mr1 2 + mr2 2 + mr3 2 )2 e2 (m1 r1 + m2 r2 + m3 r3 )r2 e2
+ (mr1 2 + mr2 2 + mr3 2 )3 e3 (m1 r1 + m2 r2 + m3 r3 )r3 e3
= m(r2 2 + r3 2 )1 e1 m(r1 r2 )2 e1 m(r1 r3 )3 e1
m(r2 r1 )1 e2 + m(r1 2 + r3 2 )2 e2 m(r2 r3 )3 e2
m(r3 r1 )1 e3 m(r3 r2 )2 e3 + m(r1 2 + r2 2 )3 e3
Similar to the rotation of a vector, the components of the angular momentum and angular
velocity vectors can be expressed as column vectors and the inertia in a matrix form as



h1
m(r2 2 + r3 2 ) m(r1 r2 )
m(r1 r3 )
1
I11 I12 I13
1



2
2
h2 = m(r2 r1 ) m(r1 + r3 ) m(r2 r3 ) 2 = I21 I22 I23 2
h3
m(r3 r1 )
m(r3 r2 ) m(r1 2 + r2 2 ) 3
I31 I32 I33
3
At first glance, the structure of the above equation looks very similar to that of transformation of vectors between bases. However, the inertia matrix, I and the transformation
matrix, A represent two fundamentally different quantities. The transformation matrix
operates on a vector to result its components in a new basis. Note that the composition of
the transformation matrix is independent of the vector it operates on i.e., the components
aij are not derived from the components of the vector they act on, but are derived from the
two bases involved. The inertia matrix, however, is explicitly derived from the position
of the particle (r) and its mass. Consequently, any change in r is summarily reflected in
14

I. Thus, I has a more sound physical basis and significance than A and represents a property/ state of the particle in the current basis, in addition to performing the mathematical
function of mapping to h. Such quanities are called as Tensors. However, this does not
imply that A is not a tensor. Indeed it is one and belongs to a class of orthogonal tensors.
We have presented arguments that the components of the Inertia Tensor are derived
from components of the position vector, r. This implies that if one changes the basis such
that the components of r are altered, then the components of I should also change, i.e.,
the inertia tensor (and tensors in general) are defined in a specific basis, similar to vectors.
Further we note that each component of I is associated with two basis vectors.
For example, consider how the second row of the I matrix operates. We have h2 =
I21 1 + I22 2 + I23 3 . We know that h2 represents the component of the angular momentum in the e2 direction and 3 represents the component of the angular velocity in the e3
direction. The component I23 is thus mapping a component along e3 to a quantity along
e2 , i.e., each component of I is associated with two basis vectors. It may be argued that
I22 deals exclusively with the e2 direction. The relevant interpretation here is that I22
maps one component of the vector to one component of the vector h - they just happen
to be along the same direction! Thus, each component of the tensor I is associated with
two basis vectors and consequently, would require two basis vectors to describe/ represent
it.
Also evident is that a vector, which is associated with one basis vector for each component, can be represented by a column, whereas I requires a two dimensional structure
(matrix) for its representation since each component of I is associated with two basis
vectors. Thus, we refer to I as a second order tensor. Similarly, vectors are referred to as
first order tensors and scalars are zero order tensors. Since we have identified inertia to
be a second order tensor, we represent it as I.
We have previously seen operations such as the dot product or the scalar product in
which a vector is reduced to a scalar, i.e., a first order tensor is reduced to a zero order
tensor. Here, we introduce the notion of tensor product or dyadic product represented
by the symbol. In this operation, two first order tensors (vectors) are operated on to
form a second order tensor given by ei ej = ei ej . The dyadic or tensor product of two
vectors can be written as
u v = ui ei vj ej = ui vj ei ej = ui vj ei ej = Tij ei ej
where Tij are the components of the second order tensor T formed by the tensor product
of u and v. The second order tensor has two free indices represented by the basis vectors
ei and ej and its components are represented as a matrix. In the present case, the the
15

inertia tensor can now be written as


Iij = m(rk rk ij ri rj )
I = m[(r r)I r r]
where I is the identity tensor given by I = e1 e1 + e2 e2 + e3 e3 . Note that all
tensors need not be defined based on position vectors. There are many instances where
second and higher order tensors are used to describe a specific property of a material,
independent of position vectors.
Once the Inertia tensor has been formed, it acts upon the angular velocity vector to
result the angular momentum vector. We have addressed the fact that the Inetia tensor
is a second order tensor (I) and the angular velocity vector is a first order tensor ().
This implies that I operates on to reduce the number of free indices i.e., the operation
beween I and is a scalar or dot product. This can be written as
h=I
= Iij ei ej k ek
= Iij k ei ej ek
= Iij k ei jk
= hi ei = Iij j ei
which is consistent with previous results.
Properties of Tensors
We have addressed the properties of vectors in a previous subsection in which we
observed that vectors belong to a linear space and since they are defined using a basis,
they transform according to certain rules. The same holds good for tensors as well.
Tensors also belong to the same space. Here, we distinguish between a space containing
the collection of vectors, u, v etc. and the vector space, V. The vector space, V that
we have used thus far is not necessarily limited to the collection of all possible, physical
(R3 ) vectors. Rather, it is a more generalised space of linear operators that satisfy the
linear vector space rules outlined in the first section. Thus, all (R3 ) vectors belong
to V, but do not completely define V. The tensors we have defined can be viewed as
linear operators on vectors, and so, belong to this general space of linear operators and
satisfy all the rules associated with belonging to a linear vector space. We have seen
previously the rules governing change of vector representation with a change in the basis
and realized that quantities to be identified as vectors must transform according to these
specific rules. Since tensors are also expressed in a specific basis, there are specific rules
to be satisfied during transformation of bases. Consider the tensor I we have dealt with
16

so far in the ei basis and is transformed to I in the ei basis.


Iij ei ej = Ikl ek el
= Ikl amk em anl en
= Ikl amk anl em en
= amk Ikl (aln )T em en
= I = AIAT

(1.12)

We note from the definition of the tensor I that when represented in a matrix, I is
symmetric, i.e., Iij = Iji . If, for some tensor A, Aij = Aji then that tensor is said to be
skew-symmetric. To satisfy the requirements of skew-symmetry, the diagonal elements of
such a tensor are 0. In general, it is not necessary for a tensor to satisfy the requirements
of symmetry or skew-symmetry. However, any tensor can be represented as a sum of a
symmetric and a skew-symmetric tensor, i.e., any tensor C can be written as
C=A+B
Aij = Aji is a symmetric tensor, where A =

C + CT
2

Bij = Bji is a skew-symmetric tensor, where B =

C CT
2

We have seen that the Inertia tensor maps the angular velocity vector into the angular
momentum vector. We summarize that the angular momentum is defined about a point
that is considered to be the origin and the angular velocity is a vector that passes through
this origin. The position vectors are also defined with respect to the origin. Also, the
Inertia tensor maps the angular velocity vector to the angular momentum vector such
that they need not be along the same direction. The formula for the angular momentum
is given by
h = mr v = m(r r) m(r )r
The angular momentum vector, h is comprised of two different parts, one along and
the other along r. It is obvious that when r, the part of h along r goes to 0 and hk.
We can present these arguments about the angle between and r and the associated
orientation of h for the idealized example we have considered, i.e., a point mass which is
described by a single r.
For a finite dimensional body, i.e., something which cannot be approximated to a point
mass and thus cannot be described by a unique position vector r, the above generalizations
would not hold. The Inertia tensor for such a body would have to be obtained by
considering the moments of inertia of infinitismal elements of the body and integrating
them. For such an object, the only mathematical possibility for h to lie along is for
17

I to be a purely diagonal matrix, i.e., the off-diagonal terms should vanish. For each
infinitismal elemet considered, the components of the Inertia tensor of that element are
functions of the components of the position vector of that infinitismal element, i.e., I12 is
a function of r1 and r2 , I13 is a function of r1 and r3 etc. We recall that the components
of any vector depend on the basis in which they are described. Thus, one may select
a basis to represent the finite dimensional body such that the off-diagonal components
of the Inertia tensor for a given infinitismal element may be non-zero, but when that
component of the inertia tensor is integrated over the domain of the body, it reduces to
zero. In such a case, the inertia tensor for the entire body would be represented by a
diagonal matrix and consequently, h k . Mathematically, this condition can be written
as
h = I =
I = 0
I I = 0
(I I) = 0
We know that the solution for this exists when




I I = 0

(1.13)

(1.14)

The set of vectors which from the basis in which this condition is satisfied are called as
Eigenvectors and the values of thus obtained are called as Eigenvalues. We can
solve for these by expressing the above equation in a matrix form as




I I = 0




= Iij Iij = 0


I11

0
0




I22
0 =0
0


0
0
I33
which results in a cubic equation in i.e., three possible Eigenvalues, I , II and III .
For each of these Eigenvalues, there exixts a vector which satisfies Eqn. (1.13) and is the
corresponding Eigenvector, i.e., for = I , there exists a vector I such that
(I I I) I = 0
The set of vectors for all the three possible values of are called as the Eigenvectors.
Physically, this means that around this set of vectors/ axes the mass distribution is symmetric. For a tensor whose components are real valued and symmetric, the Eigenvectors
form an orthogonal triad.

18

1.3

Calculus of vectors and tensors

Thus far, we have addressed the algebra of vectors, which have been primarily defined at
a specific point at a specific instant of time. We have not considered cases of quantities
varying either spatially or temporally. In most physical situations, quantities do vary
with both space and time. Here we introduce the relevant mathematics to quantify these
variations.
Consider a particle that moves with progression of time i.e., the position vector of
the particle changes with time and can thus be represented as a function of time as r =
r(t). Let S be the path that the particle traces. We consider the particle at two instances
of time, t and t + t. Then the velocity of the particle is given by the derivate of the
particle position as
r
v = lim
t0 t

Figure 1.7: Differentiation of the position vector.


If there are no discontinuities in the path S, then,

r
r
S
=
t
S t


From the figure, it can be seen that as t 0, r S and r/S et where et
is the tangential vector to the path S. Thus,
v = lim et
t0

S
dS
=
et
t
dt

i.e., the magnitude of the velocity vector is the rate at which distance is covered, but it
is oriented tangential to the path along which the particle is travelling. If we express et
in terms of the basis vectors ei , then we can express the velocity as v = vi ei . Note that
since ei s are independent of time, t, this implies that the velocity can be integrated to
obtain a general expression for the position of the particle as r = ri (t)
ei .

19

Gradient operator
Quite often in continuum description of physical properties and processes, we encounter spatial variations of properties/ parameters, meaning, these quantities can be
expressed as a function of the coordinates of a given point in space, i.e., they can be
expressed as Fields. The simplest fields to imagine and to mathematically describe are
scalar fields. Since the field describes a scalar quantity, it is not associated with any
particular direction. A typical example could be the spatial distribution of temperature.
The value of the temperature field at any given point in space is a function of the components of the position vector of that point. The general form of such a scalar valued
function at a point P, is given by
= (x, y, z) = (x1 , x2 , x3 )
e1 +y
e2 +z
e3 x1 e1 +x2 e2 +x3 e3 .
Note that the position vector of point P is given by P = x
has been defined to be a continuous function dependent on all the three coordinates.
Thus, the total differential for is given by

x +
y +
z
x
y
z

=
x1 +
x2 +
x3
x1
x2
x3

xi
=
xi

We note that the xi s are essentially components of an infinitismal vector, n, i.e., n =


x1 e1 + x2 e2 + x3 e3 . We also note that the terms involving the partial derivatives could
possibly have directional dependence. We know that is a scalar valued function, i.e., it

, one finds the


does not have components in specific directions. However, to evaluate
x1
rate of change of only as x1 changes, which would obviously be along e1 . Thus, the set
of partial derivatives evaluated are actually direction dependent and represent the rate
of change of along the directions of the basis vectors. The above equation for change
of can then be written as

x1 +
x2 +
x3
x1
x2
x3



=
e1 +
e2 +
e3 (x1 e1 + x2 e2 + x3 e3 )
x1
x2
x3

The function is a scalar function and does not have any components, implying that the
directional dependence in the above relation is derived from the operator determining the
partial derivative of . This operator is represented using the symbol and referred to

as the Del or the grad or the nabla operator. Mathematically, it represents the Gradient
of . Thus,

=
e1 +
e2 +
e3 such that =
e1 +
e2 +
e3 =
ei (1.15)

x1
x2
x3
x1
x2
x3
xi
20

The gradient of is rate at which the scalar function changes in space. The expression
for indicates that the total change in is equal to the rate of change of dotted
with the direction of interest. More specifically, the above equation would represent the
change of along a specific direction n given by
n = n
(1.16)

where n has not been specified to be a unit vector. From the definition of the dot
product, the value of the the above dot product is maximum when n k . This means

that is maximum along the direction of , i.e., the gradient of a field is a vector

that gives the magnitude and direction of the greatest change for that field. The vector
where S is the length of the
n can be written in terms of a unit vector as n = S n
vector n. Then, Eqn. (1.16) can be written as
d
n = S n
=
= n

(1.17)

dS
where d/dS is the rate of change of along n
. Of special interest is the case when
we find encounter spatially constant fields. Let us say that that there exists a field
(x, y, z) = constant which represents a level surface. This means that if one moves
along the surface of this field , there would be no change in its value. Now, in Eqn.
(1.16), let us choose n to be a unit vector that lies along the surface of this plane, i.e., in
the tangential direction, t. Since is constant along the horizontal surface, we can write
Eqn. (1.16) as
t = t = 0
(1.18)

This has to be valid for any t chosen to lie in the plane of (x, y, z) = constant, i.e.,

has to be normal to the surface (x, y, z) = constant. Physically, we are implying that
the direction of the greatest slope of a plane surface is normal to it! Does this make sense?
From Eqn. (1.15), we see that is a directional operator i.e., its components are

associated with specific basis vectors. Further, the gradient operation of a scalar quantity,
yields a vector, i.e., there is no contraction of bases and the operation is similar to the

dyadic product outlined earlier. Thus, the gradient of a vector u would be represented as



u=
e1 +
e2 +
e3 (u1 e1 + u2 e2 + u3 e3 )

x1
x2
x3

=
ei uj ej
xi
uj
=
ei ej
xi
uj
= u =
ei ej
(1.19)

xi
The operator acts on a vector to result a second order tensor. We will see examples

and applications of this later on in the course.

21

Divergence
In the previous section, we have seen that vectors can be involved in operations like
dot products, cross products and dyadic product. The del operator ( ) has been shown

to be a vector and the gradient operation essentially forms a tensor product. We now
focus on the scalar or dot product using the del operator. Of course, a vector can be
dotted only with another vector or a tensor of higher order, but not a scalar. A simple
example of a vector field is shown in the figure below, which depicts the velocity variation
in an idealized fluid flow (recall from ME 202!). We recall that the general form of such
a field is
e1 + v2 (x, y, z)
e2 + v3 (x, y, z)
e3
(1.20)
v(x, y, z) = v1 (x, y, z)
and note that any component of the vector field can be dependent on all the three
coordinates at a given point.

Figure 1.8: Example of the velocity field of fluid flow.


We define the Divergence of the vector field v, represented by div v as
div v = v




=
ei vj ej
xi
vj
=
ei ej
xi
vj
=
ij
xi
vi
= div v = v =

xi

(1.21)

The form of the above equation indicates that the divergence of v is a scalar, since no
free indices are present. Further, the divergence of v is a summation over the repeated
index i, given by
vi
v1
v2
v3
div v =
=
+
+
(1.22)
xi
x1 x2 x3
Physically, the divergence of a vector field describes and measures the propensity of a
field to radially converge or diverge at a point in space. Concomitant to the physical
meaning of divergence of a vector field, it can be quickly verified that for the simplistic
example of fluid flow presented earlier, the divergence of the field is 0 at all points in space.

22

The implication of the divergence operator is elucidated in greater detail in later sections.
Curl
The other vector algebraic operation is the vector or the cross product. When the del
operator () acts on a vector field similar to the cross product, the result is called as the

Curl of the vector field v. Recall that the vector product of two vectors also results in a
vector. This is given by
curl v = v


e1
e2
e3






=

x1 x2 x3


v1
v2
v3






v3
v3
v2
v2
v1
v1
=
e1
e2 +
e3

x2 x3
x1 x3
x1 x2
This can be written in indicial form as


curl v = v =
ei vj ej

xi
vj
ei ej
=
xi
vj
=
ijk ek
xi
vj
= v = ijk
ek

xi

(1.23)

The curl of a vector field is also referred to as the rotation of the vector field since it is
proportional to the angular velocity at a point in space. Again, the details are presented
in subsequent sections. A vector field for which v = 0 at all points in space is called

an irrotational field.
We introduce the following notation to simplify the representation of partial derivatives. The partial derivative of any quantity with respect to xi is represented using ,i .
Thus,

ei = ,i ei
grad = =

xi
uj
grad u = u =
ei ej = uj,i ei ej

xi
ui
div u = u =
= ui,i

xi
uj
curl u = u = ijk
ek = ijk uj,i ek

xi
23

(1.24)

Integral theorems
The most commonly used integral theorem that relates integrals of fields is called as
Gauss Theorem. However, Gauss theorem itself makes use of the result from the Greens
theorem which relates a volume integral into an integral over its bounding surface. The
Greens theorem states that, if f is a scalar field having continuous partial derivaties for
each of the cartesian basis ei , then
Z
Z
f
dV = f ni dS
xi
S

The theorem is also applicable in lower dimensions, i.e., an integral over a surface S can
be transformed to a contour C as
Z
Z
f
dS = f ni dC
xi
S

From the above integrals, it is obvious that we can define f to be components of a


vector field v and the sum of the integrals over all components would yield
Z
Z
vi
dV = vi ni dS
xi
V
S
(1.25)
Z
Z
v=

=
V

vn

The above result is called the Divergence theorem or Gauss theorem which states
that the divergence of a vector field over a volume is equal to the component of the vector
along the outward normal integrated over the surface that bounds the volume, as shown
in Fig. 1.3. We do have constraints that the vector field and its derivatives are continuous
and that the surface is closed.

Figure 1.9: Transformation of a volume integral into a surface integral - Gauss Theorem
The Gauss theorem can also be expressed in a more general form as
Z
Z
? A dV = n
? A dS

24

(1.26)

where A could represent a scalar, vector or a tensor field and ? represents operations on
the fields such as the dot product, vector product or the dyadic product.
Another important integral transformation theorem relates the integral over a surface,
S to the integral over a closed contour, C, shown in Fig. 1.3. We do need to follow certain
conventions while defining the orientations. The tangent vector, t along the contour C is
defined to conform to the right hand sense with the thumb pointing towards the direction
of the normal, n
of the surface S.

Figure 1.10: Transformation of a surface integral into a contour integral - Stokes Theorem
The general version of the Stokes theorem can then be written using A and ? defined
above, as
Z
Z
(
n ) ? A dS = t ? A dC
(1.27)

For the specific case of A representing a vector field v and ? representing the dot product,
we get
Z
Z
( v) n
dS = t v dC
(1.28)

25

También podría gustarte