You are on page 1of 10

NPTEL – Mechanical Engineering – Continuum Mechanics

Module-1: Tensor Algebra


Lecture-2: Vector and Vector Spaces

We develop the tensor algebra that is required to study the continuum mechanics. In this
lecture, we introduce the concept of vector, vector space, basis, and inner product.
Elementary concept of vector
The vector in elementary physics is defined as the quantity which has magnitude and
direction. The vector is graphically represented by an arrow shown in Fig. (1a) while
the algebraic representation (in 3-D space) is done with three numbers. For example,
vector u is represented by {u1 , u2 , u3 } or u1 e1 + u2 e2 + u3 e3 , q
where e1 , e2 , e3 are unit
vectors along coordinate axes. The length of vector, |u| = u21 + u22 + u23 , represents
the magnitude while direction cosines (cos θ1 , cos θ2 , cos θ3 ) represent the direction of
vector as shown in Fig. (1a). Two vectors can be added and subtracted as shown in
Figs. (1b) and (1c). Two vectors are said to be equal if they have same magnitude and
direction. Examples for the vector quantities are position, velocity, acceleration etc. It
can be observed that the coordinate transformation can alter the components but does
not affect the vector. For example, somebody throws a stone then the velocity vector of
stone does not depend on the choice of coordinate system, i.e., the existence of the vector
is independent of the coordinate frame. However, the components of velocity vector do
depend on the coordinate frame, i.e., the column matrix that represents the velocity vector
does change with coordinate frame. The foregoing discussion on vectors is intuitive from
the elementary physics.

1_
2
|u| = (u12 + u22 + u32 )

cos θ1 = |_
u1
u|
u+ v y u = y-x
u cos θ2 = |u_2
u|
v
θ2
cos θ3 = |_
u3
u|
θ1 u x
θ3

Figure 1: Vector in Euclidean three-dimensional space: (a) vector representation, (b)


vector addition (c) vector subtraction

It is clear from the elementary definition of vector that unit vectors along coordinate
axes are used without really defining them. Since unit vectors along coordinate axes
themselvs are vectors, they cannot exist without having definition of vector. Clearly,

Joint initiative of IITs and IISc – Funded by MHRD, Government of India 1


NPTEL – Mechanical Engineering – Continuum Mechanics

elementary definition of vector cannot be done without unit vectors along coordinate axes
and also unit vectors along coordinate axes cannot stand without defining vector. This
is running into the egg and chicken paradox. In order to resolve the paradox, one should
follow axiomtic framework. We now define vector space and vectors by axioms. Later,
the unit vectors along coordinate axes can be recoverd by the definition of basis.
Vector space
We consider only the vector spaces over the field of real numbers. Let us denote the set
of real numbers by < and vector space by V. The set, V, equipped with an addition
operation (V × V → V) and a scalar multiplication operation (< × V → V), is called
vector space if it follows:
(i) Associativity: (u + v) + w = u + (v + w), for all u, v, w ∈ V.
(ii) Commutativity: u + v = v + u, for all u, v ∈ V.
(iii) Existence of a zero element: There exist 0 ∈ V such that u + 0 = u, for all u ∈ V.
(iv) Existence of negative elements: For each u ∈ V there exist a unique v ∈ V such that
u + v = 0 and v is denoted as −u.
(v) Associativity in scalar multiplication (or the compatibility of multiplication defined
between the field elements (real numbers) and scalar multiplication operation): α(βu) =
(αβ)u, for all u, ∈ V, and α, β ∈ <.
(vi) Identity in scalar multiplication: There exist a unique element 1 ∈ < such that
1u = u, for all u ∈ V.
(vii) Distributivity with respect to addition operation on vectors: α(u + v) = αu + αv,
for all u, v ∈ V, and α ∈ <.
(viii) Distributivity with respect to scalar addition: (α + β)u = αu + βu, for all u ∈ V,
and α, β ∈ <.
The elements of set V are called vectors. This axiomatic framework generalizes the
notion of vector. We now present few examples for the vector space.
Examples of vector spaces
Example-1: Let us consider the set of all n-tuples,

<n = {(u1 , u2 , · · · , un ) : ui ∈ <}

Addition operation:
u + v = (u1 + v1 , u2 + v2 , · · · , un + vn )
Scalar multiplication operation:

αu = (αu1 , αu2 , · · · , αun )

It is easy to verify that the set, <n , obey all the axioms of vector space under the addition
and scalar multiplication. Therefore, the set, <n , is a vector space. It can be observed
that the special case n = 3 is our usual three-dimensional (3-D) Euclidean space.

Example-2: Let us consider the set of all m × n matrices over the real numbers,
 
· · · a1n


 a11 a12 

· · · a2n
 

 a21 a22  

<m×n =
 
 .. .. .. ..  : aij ∈ <




 . . . .

 


 
am1 am2 · · · amn
 

Joint initiative of IITs and IISc – Funded by MHRD, Government of India 2


NPTEL – Mechanical Engineering – Continuum Mechanics

Addition operation:
···
 
a11 + b11 a12 + b12 a1n + b1n

 a21 + b21 a22 + b22 ··· a2n + b2n 

A+B =  .. .. .. .. 

 . . . .


am1 + bm1 am2 + bm2 · · · amn + bmn
Scalar multiplication:
· · · αa1n
 
αa11 αa12

 αa21 αa22 · · · αa2n 

αA =  .. .. .. .. 

 . . . .


αam1 αam2 · · · αamn
It is easy to verify that the given set, <m×n , with defined operations is a vector space.
Example-3: Let us consider set of all Lebesgue measurable functions over the domain of
[0,1], as stated in the following expression,
 Z 1 
p p
L [0, 1] = f : |f | dx < ∞ ,
0

where p is a positive real number.


Addition operation:
(f + g)(x) = f (x) + g(x), x ∈ [0, 1]
Scalar multiplication:
(αf )(x) = α(f (x)), x ∈ [0, 1]
It is also easy to verify that the set, Lp [0, 1], is a vector space over the defined operations
and this vector space is known as Lp -space.
From these examples, it is clear that the definition of vector space is not limited to the
2-D or 3-D Euclidean spaces. This general definition of vector space has many advantages
in the study of linear algebra and operator theory. We now state the formal definition of
coordinate axes or coordinate frame.
Coordinate frame and basis:
We know from the discussion on elementary concept of vector that the definition of coordi-
nate vectors are useful in the representation of other vectors. Therefore, this necessitates
the definition of a basis which acts as coordinate frame. We can visualize the coordinate
axes as the geometric lines/curves along the basis vectors. The definition of basis require
set of linearly independent vector which is stated next.

Linear independence:
A subset {u1 , u2 , · · · , un } of V is said to be linearly dependent if and only if there exist
set of scalars α1 , α2 , · · · , αn , not all zero, such that

α1 u 1 + α2 u 2 + · · · + αn u n = 0

If such nonzero scalars do not exist, i.e, α1 = α2 = · · · = αn = 0, then the set vectors,
{u1 , u2 , · · · , un }, are said to be linearly independent.
Basis vectors:
A subset {u1 , u2 , · · · , un } of V is said to be the basis if the subset is linearly indepen-
dent and linear combination of this subset spans the total set V, i.e. there exist some

Joint initiative of IITs and IISc – Funded by MHRD, Government of India 3


NPTEL – Mechanical Engineering – Continuum Mechanics

α1 , α2 , · · · , αn ∈ <, such that x = α1 u1 +α2 u2 +· · ·+αn un for every x ∈ V. The number of


vectors in the subset is called dimension of vector space. The basis vectors are the tangents
to the coordinate axes. It is clear that the basis vectors are not necessarily orthogonal and
also not unique but linearly independent. For example, {(1, 0, 0), (0, 1, 0), (0, 0, 1)} is one
set of orthogonal basis to 3-D Euclidean space whereas the set {(1, 0, 0), (1, 1, 0), (1, 1, 1)}
is also another basis but non-orthogonal. Every vector in 3-D space can be generated by
the linear combination from any one of the set of basis vectors.

Length of the vector:


None of the axioms of vector space gives neither an idea of length of the vector nor angle
between vectors. But, we require length of vector and angle between vectors to study
the quantities such as strain and strain rates in continuum mechanics. We know from
the elementary physics that the dot product involves length of vectors as well as angle
between vectors. Therefore, we introduce the length of vector through the dot product.
Dot product on 3-D Euclidean space
Let us consider 3-D Euclidean space with canonical basis {e1 , e2 , e3 }, i.e., e1 = (1, 0, 0),
e2 = (0, 1, 1) and e3 = (0, 0, 1). Any vector in this space can have three components along
{e1 , e2 , e3 } directions. For example the vector u is represented as either (u1 , u2 , u3 ) or
u1 e1 + u2 e2 + u3 e3 with the meaning of u1 , u2 , u3 are components of u along {e1 , e2 , e3 }
directions (see Fig. (2a)). We now recall the definition of dot product (in 3-D space) from
the elementary physics,
u · v = |u||v| cos(θ), (1)
where |u| and |v| are length of vectors u, v, respectively and θ is the angle between the
vectors u and v as shown in Fig. (2b).

Figure 2: Vector representation in 3-D: (a) components of vector (b) two vectors separated
by an angle θ.

The length of the vector is defined by

u · u = |u|2 = u21 + u22 + u23 (2)

Now, an alternative definition is stated for dot product which is equivalent to the one
defined in Eq. (1).
u · v = u1 v1 + u2 v2 + u3 v3 (3)
We note that the definition of dot product given in Eq. (3) accounts for both length of
vector shown in Eq. (2) and also angle between vectors shown in Eq. (1). We will prove
the equivalence of definitions stated in Eq. (1) and Eq. (3).

Joint initiative of IITs and IISc – Funded by MHRD, Government of India 4


NPTEL – Mechanical Engineering – Continuum Mechanics

Problem 1. In 2-D Euclidean space with canonical basis ({e1 = (1, 0), e2 = (0, 1)}) the
definition of dot product defined in Eq. (1) and Eq. (3) are equivalent, i.e. |u||v| cos(θ) =
u1 v1 + u2 v2

Proof-1. Let us consider two vectors in two dimensional space as shown in the Fig. (3).
We note that e1 is unit vector along horizontal axis and e2 is unit vector along vertical
axis.

u2 u

v
v2 θ
θv θu
u1
v1

Figure 3: Two vectors in 2-D space

The length of vectors can be represented by


1 1
|u| = (u21 + u22 ) 2 , |v| = (v12 + v22 ) 2 . (4)

Let θu and θv be angles made by vectors u and v with respect to horizontal axis as shown
in Fig. (3). Therefore, we have
u1 u2 v1 v2
cosθu = , sinθu = , cosθv = , sinθv = . (5)
|u| |u| |v| |v|

Combining these relations, we get


u1 v1 u2 v2
+ = cosθu cosθv + sinθu sinθv = cos(θu − θv ) (6)
|u| |v| |u| |v|

It is easy to observe from Fig. (3) that θ = θu − θv . Therefore, we have

u1 v1 + u2 v2 = |u||v|cosθ (7)

The result can be generalized to 3-D as any two vectors from two-dimensional subspace.
We can have the following alternative proof for the same result based on cosine rule.
Proof-2. We have two vectors u and v with an angle θ between them. Therefore, the
vectors u, v and u − v forms
q a triangle
q as shown q
in Fig. (4). Three sides of triangle are
2 2 2 2
|u|, |v| and |u − v|, i.e., u1 + u2 , v1 + v2 and (u1 − v1 )2 + (u2 − v2 )2 .

Joint initiative of IITs and IISc – Funded by MHRD, Government of India 5


NPTEL – Mechanical Engineering – Continuum Mechanics

u u-v

θ
v

Figure 4: Two vectors with an angle θ

We now apply the cosine rule to the triangle shown in Fig. (4), i.e.,
|u − v|2 = |u|2 + |v|2 − 2|u||v|cosθ (8)
We can rewrite Eq. (8) as
1
|u||v|cosθ = (|u|2 + |v|2 − |u − v|2 )
2
1 2 1 1
= (u1 + u22 ) + (v12 + v22 ) − ((u1 − v1 )2 + (u2 − v2 )2 )
2 2 2
= u1 v1 + u2 v2
Thus, the result is proved. This can be easily generalized to the vectors in 3-D space.
Inner product:
We saw the general definition of vector space in the previous discussion. Now, we gener-
alize the definition of dot product over the vector space, which is known as inner product
or scalar product.
Let V be a vector space. The inner product is a function from V × V → <, denoted by
(u, v), satisfying the following properties:
(i) Linearity: (αu + βv, w) = α(u, w) + β(v, w) ∀α, β ∈ < and ∀u, v, w ∈ V
(ii) Symmetry: (u, v) = (v, u) ∀u, v ∈ V
(iii) Positive-definiteness: (u, u) ≥ 0, ∀u ∈ V and (u, u) = 0 if and only if u = 0.
The definition of inner product brings out the geometric quantities such as length and
orthogonality. The vector space equipped with inner product is called inner product space
or Hilbert space.
Example-4: There is an inner product on the vector space <n . It is defined as,
n
X
u · v = u1 v1 + u2 v2 + · · · + un vn = ui vi (9)
i=1

This definition satisfies all the properties of inner product. It is known as standard inner
product on <n .
Example-5: The choice of inner product is not unique. For any given positive definite
matrix S, another choice of inner product on <n is
n X
X n
u·v = ui Sij vj (10)
i=1 j=1

Joint initiative of IITs and IISc – Funded by MHRD, Government of India 6


NPTEL – Mechanical Engineering – Continuum Mechanics

where Sij is ij th component of S.


This choice of inner product has application in measuring length of vector and angle
between two vectors when the basis of <n are non-orthogonal. The application of inner
product involving positive definite matrix is explained through the following numerical
example.
Example-6: Let us consider two vectors, u and v, in 2-D vector space (<2 ) as shown in Fig.
(5). As stated earlier the basis are not unique to the vector space. Hence, consider two
different basis to the vector space. Let first basis be the canonical basis, i.e. B1 = {e1 , e2 }
where e1 = (1, 0), e2 = (0, 1) and other basis be a non-orthogonal basis B2 = {e01 , e02 }
 1 1 
where e01 = (1, 0), e02 = √ , √ . It is trivial to find length of vectors and angle between
2 2
vector in canonical basis. We now present the way to find the length and angle between
vectors in non-orthogonal basis.

u = e1 + 2e2

e2 e'2 v = 2e1 + e2

e 1 e'1

Figure 5: Two vectors in the space of <2

Solution: Given vectors, u = e1 + 2e2 and v = 2e1 + e2 are in canonical basis. Length of
√ √
vectors, |u| = 5 and |v| = 5.
Let θ be the angle between two vectors. Then
u·v u1 v1 + u2 v2 4
cos θ = = =
|u||v| |u||v| 5
Both vectors u and v have the following representation in the basis B1 and B2 .

u = u1 e1 + u2 e2 = e1 + 2e2 in the basis B1



= u01 e01 + u02 e02 = −e01 + 2 2e02 in the basis B2
v = v1 e1 + v2 e2 = 2e1 + e2 in the basis B1

= v10 e01 + v20 e02 = e01 + 2e02 in the basis B2

Let us consider a positive definite matrix,


1
 
 1 √
2

S :=  1
 


√ 1

2
The inner product in the basis B2 ,
2 X
2
u0i Sij vj0
X
(u, v) =
i=1 j=1
= u01 S11 v10 + u01 S12 v20 + u02 S21 v10 + u02 S22 v20
= 4

Joint initiative of IITs and IISc – Funded by MHRD, Government of India 7


NPTEL – Mechanical Engineering – Continuum Mechanics

Length of vectors,
 1  1
q 2 X
2 2
√ q 2 X
2 2

u0i Sij u0j  = 0 0
X X
|u| = (u, u) =  5 and |v| = (v, v) =  v Sij v
i j = 5
i=1 j=1 i=1 j=1

The angle between two vectors,


(u, v) 4
cos θ = =
|u||v| 5
This example shows that a positive definite matrix is required to get physically meaningful
measure length and angle in non-orthogonal basis. Furthermore, if the positive definite
matrix is identity then we can recover usual inner product. Thus, the inner product with
respect to a positive definite matrix not only provide useful measure of length and angle
but generalizes the concept of inner product.
Example-7: The vector space L2 [0, 1] is an inner product space. The inner product over
L2 is defined by Z 1
(f, g) = f (x)g(x)dx, ∀f, g ∈ L2 [0, 1].
0
It can be observed that this definition follows the set of axioms defined for the inner
product space.

Orthogonality:
If the inner product between two non-zero vectors is zero then two vectors are said to be
orthogonal. Let u and v be orthogonal vectors in an inner product space V. Then

(u, v) = 0.

In two-dimensional or three-dimensional spaces if the vectors are orthogonal then the


angle between them would be π/2. Thus, the inner product brings out the concept of
orthogonality. Furthermore, vectors in inner product space follow an important inequality
called Cauchy-Schwartz inequality.
Cauchy-Schwartz inequality:
Let V be an inner product space. Then

(u, v)2 ≤ (u, u)(v, v), ∀u, v ∈ V (11)

and the equality holds if and only if u and v are linearly dependent.

Proof. If (u, v) = 0 then the inequality is trivial. Let us assume both vectors u and v
are non-zero. The positive definite property of inner product implies

f (α) = (u − αv, u − αv) ≥ 0 ∀α ∈ <

Using linearity, i.e., property (i) of inner product, we get

f (α) = (u, u) − 2α(u, v) + α2 (v, v) ≥ 0

The quadratic function f (α) is minimum at α = (u, v)/(v, v). Thus, we obtain the
Cauchy-Schwartz inequality by substituting α = (u, v)/(v, v) in previous equation. If
equality holds in Eq. (11) then we get (u − αv, u − αv) = 0 where α = (u, v)/(v, v). The
relation (u − αv, u − αv) = 0 implies u = αv, i.e., u and v are linearly dependent.

Joint initiative of IITs and IISc – Funded by MHRD, Government of India 8


NPTEL – Mechanical Engineering – Continuum Mechanics

The application of the Cauchy-Schwartz inequality to <n and L2 [0, 1] spaces yields

n
!2 n
! n
!
u2i vi2 ,
X X X
ui vi ≤
i=1 i=1 i=1
Z 1 2 Z 1  Z 1 
f (x)g(x)dx ≤ f (x)2 dx g(x)2 dx .
0 0 0

The triangle inequality follows from the Cauchy-Schwartz inequality. The proof is pre-
sented next.

Triangle inequality:
Let V be an inner product space. Then

|u + v| ≤ |u| + |v| ∀u, v ∈ V

Proof:

|u + v|2 = (u + v, u + v)
= |u|2 + 2(u, v) + |v|2
≤ |u|2 + 2|(u, v)| + |v|2 (Since (u, v) ≤ |(u, v)|)
≤ |u|2 + 2|u||v| + |v|2 (By Cauchy-Schwartz inequality)
2
≤ (|u| + |v|)

We can get the triangle inequality by taking square-root on both sides. The triangle
inequality is essential in defining the normed vector space where the definition of length
or norm of vector is the main task.

Normed vector space:


Let V be a vector space. If there is a function from V → <, denoted by kuk where u ∈ V,
obey the following properties then V is called normed vector space.
(i) kuk ≥ 0 ∀u ∈ V with kuk = 0 ⇐⇒ u = 0
(ii) kαuk = |α|kuk ∀α ∈ < and ∀u ∈ V
(iii) ku + vk ≤ kuk + kvk ∀u, v ∈ V
It is clear that all inner product spaces are normed vector spaces equipped with the norm

1
kuk := |u| = (u, u) 2 .

Converse is not true, i.e., normed vector space is not necessarily an inner product space.
Consequently, the normed space is subset of general vector space and the inner product
space is a subset of normed vector space. This fact is depicted in Fig. (6). Furthermore,
the concept of length is defined for the normed space whereas the length and also angle
between vectors is defined for the inner product spaces. In continuum mechanics, we
require the definition of both length of vector and angle between vectors. Thus, all the
analysis are being done in the inner product spaces.

Joint initiative of IITs and IISc – Funded by MHRD, Government of India 9


NPTEL – Mechanical Engineering – Continuum Mechanics

Normed space Vector space

Inner product space

Figure 6: Venn diagram for vector spaces, normed spaces, and inner product spaces

Reference

1. C. S. Jog, Continuum Mechanics: Foundations and Applications of Mechanics, Vol-


ume I, Third edition, 2015, Cambridge University Press.

Joint initiative of IITs and IISc – Funded by MHRD, Government of India 10

You might also like