1

LECTURE NOTES ON
LINEAR ALGEBRA

B. O. Onasanya

University of Ibadan.
Copyright © 2024 by B. O. Onasanya
All rights reserved.
 Certification

I certify that .

i
 Dedication

We dedicate this work to My wife

ii
 Acknowledgements

We...

iii
 Abstract

This project work presents

iv
TABLE OF CONTENTS

Title Page 1
Certification i
Dedication ii
Aknowledgement iii
Abstract iv
1 MATRICES AND OTHER CONCEPTS 7
1.1 Matrices . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
1.1.1 What is a matrix . . . . . . . . . . . . . . . . . . . . . . . . . . 7
1.1.2 Some Operations on Matrices . . . . . . . . . . . . . . . . . . . 8
1.1.3 Cofactors of a square matrix . . . . . . . . . . . . . . . . . . . . 9
1.2 Elementary operations on Matrices . . . . . . . . . . . . . . . . . . . . 9
1.2.1 Elementary matrices . . . . . . . . . . . . . . . . . . . . . . . . 9
1.2.2 Row reduced echelon matrix . . . . . . . . . . . . . . . . . . . . 10
2 PERMUTATION AND/OR SYMMETRIC GROUP 12
2.1 Permutation Group and Its Operation . . . . . . . . . . . . . . . . . . 12
2.2 Polynomial in Relation to Permutation . . . . . . . . . . . . . . . . . . 15
3 DETERMINANTS 16
3.1 Determinant of a square matrix . . . . . . . . . . . . . . . . . . . . . . 16
3.2 Determinant in Relation to Symmetric Group . . . . . . . . . . . . . . 17
3.3 Some properties of determinants . . . . . . . . . . . . . . . . . . . . . . 18
3.4 Inverse of a Matrix . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
3.4.1 Inverse by elementary operations . . . . . . . . . . . . . . . . . 21
4 SYSTEM OF LINEAR EQUATIONS 22
4.1 System of Equations . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
4.1.1 Describing a System of equations . . . . . . . . . . . . . . . . . 22
4.1.2 Augmented matrix . . . . . . . . . . . . . . . . . . . . . . . . . 23

5
 4.2 Solution to System of Equations . . . . . . . . . . . . . . . . . . . . . . 24
4.2.1 Solving homogenous linear systems . . . . . . . . . . . . . . . . 24
4.2.2 Solving non-homogenous linear systems . . . . . . . . . . . . . . 25
5 EIGENVALUES AND EIGENVECTORS 29
5.1 Eigenvalues . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
5.2 Eigenvectors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
6 CAYLEY-HAMILTON THEOREM AND MINIMAL POLYNOMIAL 34
6.1 Cayley-Hamilton Theorem . . . . . . . . . . . . . . . . . . . . . . . . . 34
6.1.1 Minimal Polynomial . . . . . . . . . . . . . . . . . . . . . . . . 36
7 SIMILAR, DIAGONAL AND TRIANGULAR MATRICES 41
7.1 Similar Matrices . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41
8 SIMILAR AND TRIANGULAR MATRICES 44
8.1 Similar Matrices . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44
8.2 Triangular Matrices . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45
8.2.1 Upper and lower triangular matrices . . . . . . . . . . . . . . . 45
8.3 Reducing a System to Triangular Matrix . . . . . . . . . . . . . . . . . 47
9 VECTOR SPACES AND LINEAR TRANSFORMATIONS 49
9.1 Vector Spaces . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49
9.2 Some Results in Vector Space . . . . . . . . . . . . . . . . . . . . . . . 52
9.3 Linear Mappings/Transformations . . . . . . . . . . . . . . . . . . . . . 52
9.4 Change of Bases . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54
10 BILINEAR, QUADRATIC AND CANONICAL FORMS 60
10.1 Bilinear Form . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 60
10.2 Orthogonal Diagonalization of a Real Symmetric Matrix . . . . . . . . 64
10.3 Quadratic Forms . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66
10.4 Canonical Forms . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 67

6
Chapter 1

MATRICES AND OTHER

CONCEPTS

1.1 Matrices

1.1.1 What is a matrix

A matrix is a rectangular array of numbers from a field. Let A be a matrix of dimension
m × n,  
a11 a12 · · · a1n
.. 
 
 a21 · · · · · ·

. 
A=  .. .. .. .. 

 . . . . 
 
am1 · · · · · · amn

Remark 1.1.1 If m = n, A is a square matrix

If m = 1 and n ≥ 2, A is a row matrix
If m ≥ 2 and n = 2, A is a column matrix
For A = (aij ), the transpose of A is the matrix AT = (aji ).
A square matrix A is a skew-symmetric if AT = −A
A square matrix A is a symmetric if AT = A
A square matrix A is a normal if AAT = AT A
A square matrix A is a orthogonal if AAT = I = AT A

7
Remark 1.1.2 The square matrix

 0, if i ̸= j,
B = (bij ) = (1.1.1)
 α ∈ F, a field, for i = j

is diagonal matrix.
The n × n square matrix

 0, if i ̸= j,
In = (aij ) = (1.1.2)
 1, for i = j

is an identity matrix.

1.1.2 Some Operations on Matrices

Let A = (aij ) and B = (bij ) be two matrices.
i. If they are of same dimension, then, A ± B = (aij ± bij )
ii. If the column of A is the same as the row of B, say n
n
X
AB = aik bkm .
k=1

iii. In any case,

αA = (αaij )

Remark 1.1.3 Note that GLn (IR) is the set of all n × n matrices over a field IR. For
any A, B, C ∈ GLn (IR)
(i) AB ∈ GLn (IR)
(ii) A(BC) = (AB)C
(iii) In is the identity element
(iv) A−1 is the inverse of A
(vi) AB ̸= BA
Our conclusion about the set GLn (IR) under matrices multiplication should not be dif-
ficult.
Also, consider Mn (IR) as the set of all n × n matrices over a field IR. For any
A, B, C ∈ Mn (IR)

8
 (i) A + B ∈ Mn (IR)
(ii) A + (B + C) = (A + B) + C
(iii) O matrix is the identity element
(iv) −A is the inverse of A
(vi) A + B = B + A.
Our conclusion about the set Mn (IR) under matrices addition should not be difficult.

1.1.3 Cofactors of a square matrix

For  
a a a
 11 12 13 
A =  a21 a22 a23 ,
 
 
a31 a32 a33

a22 a23 a21 a23 a21 a22

A11 = , A12 = and A13 = .
a32 a33 a31 a33 a31 a32

1.2 Elementary operations on Matrices

Elementary (row or column) operations are performed on a matrix to obtain an equiv-
alent one. Here are row operations:
(i) Rij means interchange i-th and j-th rows
(ii) Ri (a) means multiply i-th row by a
(iii) Rij (α) means multiply j-th row by α and add the result to i-th row

1.2.1 Elementary matrices

Elementary matrices are obtained
 when 
elementary operations are performed on iden-
1 0 0
 
tity matrices. Consider I3 =  0 1 0 ,
 
 
0 0 1

9
  
0 0 1
 
E13 =  0 1 0  and
 
 
1 0 0
 
2 0 0
 
E1 (2)E2 (3)E3 (4) =  0 3 0 
 
 
4 0 4

1.2.2 Row reduced echelon matrix

A matrix is said to be in its row reduced echelon form if
(1) The leading entry in any non zero row is 1 and all other entries in the same column
are 0
(2) The leading entry in the (i)-th row is to the left of the leading entry in the (i+1)-th
row
(3) Zero row(s) is/are below any nonzero row

Definition 1.2.1 A matrix is in its canonical or normal form if it is in its row reduced
echelon form and the leading entry in the first row is in its first column

Example 1.2.2 Which of these is in its row reduced echelon form?

 
1 0 0 −3
 
 
 0 1 0 7 
(a) 



 0 0 1 4 
 
0 0 0 0
 
1 0 5
 
(b)  0 1 0 
 
 
0 1 1

Example 1.2.3 Find the row reduced echelon form of each of A and B and state if it
is in its canonical form.

10
  
1 0 2
 
(a) A =  2 −1 3 
 
 
4 1 8
 
0 1 3
 
(b) B =  2 1 −4 .
 
 
2 3 2
 
−7
1 0 2
 
Solution: For A it is I3 and for B it is B =  0 1 3 .
 
 
0 0 0

Remark 1.2.4 The number of non zero rows in the row reduced echelon form of a
matrix is its rank. Rank(A) = 3 and Rank(B) = 2.

11
Chapter 2

PERMUTATION AND/OR
SYMMETRIC GROUP

2.1 Permutation Group and Its Operation

A bijective mapping ρ : X → X is called a permutation.
Remark 2.1.1 This mapping only rearranges elements of X. If our permutations re-
arrange the elements of a set X of order n, there are only n! ways to do so. Hence, we
can only have n! permutation mappings. The set of all these permutations is denoted
Sn .

Example 2.1.2 Let X = {1, 2, 3} in which case n = 3 and n! = 6. We can have the
set of permutations on X denoted by S3 as
   
1 2 3 1 2 3
ϕ= =  = (132),
ϕ(1) ϕ(2) ϕ(3) 3 1 2
   
1 2 3 1 2 3
β= =  = (123),
β(1) β(2) β(3) 2 3 1
   
1 2 3 1 2 3
ρ= =  = (12),
ρ(1) ρ(2) ρ(3) 2 1 3

12
    
1 2 1 3 2 3
α= =  = (13),
α(1) α(2) α(3) 3 2 1
   
1 2 3 1 2 3
λ= =  = (23)
λ(1) λ(2) λ(3) 1 3 2
and  
1 2 3
e=  = (1).
1 2 3
S3 = {e, ρ, α, λ, β, ϕ} and |S3 | = 6 = 3!.

Remark 2.1.3 The permutation which does not change the elements of X is an identity
permutation.  
1 2 3
e= .
1 2 3

Example 2.1.4 We can compute the composition of the permutations ϕ and β, ϕβ to

mean first use β then ϕ (this is right to left and the direction must be kept unchanged).
In that case, ϕβ(1) = ϕ(2) = 1. So,
    
1 2 3 1 2 3 1 2 3
ϕβ =   =  = e.
3 1 2 2 3 1 1 2 3
    
1 2 3 1 2 3 1 2 3
βλ =   =  = ρ.
2 3 1 1 3 2 2 1 3

Remark 2.1.5 Note that composition of permutations is not commutative. Consider

that     
1 2 3 4 1 2 3 4 1 2 3 4
  = 
2 3 1 4 3 1 4 2 1 2 4 3
but     
1 2 3 4 1 2 3 4 1 2 3 4
  = .
3 1 4 2 2 3 1 4 1 4 3 2

13
 ◦ e ρ α λ β ϕ
e e ρ α λ β ϕ
ρ ρ e ϕ β λ α
α α β e ϕ ρ λ
λ λ ϕ β e α ρ
β β α λ ρ ϕ e
ϕ ϕ λ ρ α e β

Lemma 2.1.6 The set of all permutations on the set X under the composition of per-
mutations form a group.
We will illustrate this with an example.
Example 2.1.7 Let X = {1, 2, 3} in which case S3 permutations and 3! = 6. We can
have the permutations on X as S3 = {e, ρ, α, λ, β, ϕ} for
       
1 2 3 1 2 3 1 2 3 1 2 3
ϕ= β =  ρ =  α =  
3 1 2 2 3 1 2 1 3 3 2 1
 
1 2 3
λ= 
1 3 2
and  
1 2 3
e= .
1 2 3
This can be shown in a Caley’s table.
Example 2.1.8 Let G = {e, ρ, α, λ, β, ϕ} and define an operation on it by the table
below. This group is not commutative.
We can compute the composition of the permutations as:
    
1 2 3 1 2 3 1 2 3
  = .
3 1 2 2 3 1 1 2 3
    
1 2 3 1 2 3 1 2 3
  = .
2 3 1 1 3 2 2 1 3

14
2.2 Polynomial in Relation to Permutation
Let Y
Pn (x1 x2 x3 · · · xn ) = (xi − xj ) (2.2.1)
i>j

and δ ∈ Sn ,

δ̂(Pn ) = Pn (xδ(1) xδ(2) · · · xδ(n) ) (2.2.2)

Y
= (xδ(i) − xδ(j) ) (2.2.3)
i>j

Example 2.2.1 1. (i) P2 (x1 x2 ) = (x2 − x1 )

(ii) P3 (x1 x2 x3 ) = (x3 − x2 )(x3 − x1 )(x2 − x1 )
(iii) P4 (x1 x2 x3 x4 ) = (x4 − x3 )(x4 − x2 )(x4 − x1 )(x3 − x2 )(x3 − x1 )(x2 − x1 )
2. (i) If δ = (12),

δ̂(P2 ) = (xδ(2) − xδ(1) ) (2.2.4)

= (x1 − x2 ) (2.2.5)
= −P2 (2.2.6)

(ii) If δ = (132),

δ̂(P3 ) = (xδ(3) − xδ(2) )(xδ(3) − xδ(1) )(xδ(2) − xδ(1) ) (2.2.7)

= (x2 − x1 )(x2 − x3 )(x1 − x3 ) (2.2.8)
= P3 (2.2.9)

Remark 2.2.2 δ̂(Pn ) is either Pn (in which case δ is said to be even) or −Pn (in which
case δ is said to be odd). When the permutation δ is even, we write sign(δ) = +1 and
when odd, sign(δ) = −1. The following properties are obvious:
(i) sign(δ1 δ2 ) = sign(δ1 )sign(δ2 )
(ii) sign(δ −1 ) = sign(δ)

(i) sign(12) = −1
(ii) sign(123) = 1

Example 2.2.3

15
Chapter 3

DETERMINANTS

3.1 Determinant of a square matrix

Let A be an n × n matrix. The determinant of A denoted det(A) or △(A) or |A| is a
function
det : Mn (F) → F,
where Mn (F) is the set of all n × n matrices over a field F.
Definition 3.1.1 The laplace expansion of the determinant of a matrix A is
n
X n
X
i+j
det(A) = (−1) aij Aij = (−1)i+j aij Aij
j=1 i=1
 
a11 a12
Example 3.1.2 1. A =  , det(A) = a11 a22 − a12 a21
a21 a22
 
4 3
2. B =  , det(B) = (4)(1) − (3)(−2) = 4 + 6 = 10
−2 1

Example 3.1.3  
−2 −2 3
 
M = 1 1 −1 
 
 
−1 −2 2

16
 |M | = (−1)1+1 (−2)M11 + (−1)1+2 (−2)M12 + (−1)1+3 (3)M13
= −2M11 + 2M12 + 3M13
= −2(0) + 2(1) + 3(−1) .
= 0+2−3
= −1

3.2 Determinant in Relation to Symmetric Group

Given a square matrix A, alternatively,
X
det(A) = (signδ)a1δ(1) a2δ(2) · · · anδ(n)
δ∈Sn

Example 3.2.1 Find the determinant of

 
−2 −2 3
 
M = 1 1 −1  .
 
 
−1 −2 2

det(M ) = (signϕ)a1ϕ(1) a2ϕ(2) a3ϕ(3)

+ (signe)a1e(1) a2e(2) a3e(3)
+ (signρ)a1ρ(1) a2ρ(2) a3ρ(3)
+ (signα)a1α(1) a2α(2) a3α(3)
+ (signλ)a1λ(1) a2λ(2) a3λ(3)
+ (signβ)a1β(1) a2β(2) a3β(3)

det(M ) = a13 a21 a32

+ a11 a22 a33
− a12 a21 a33
− a13 a22 a31
− a11 a23 a32
+ a12 a23 a31

17
 det(M ) = (3)(1)(−2)
+ (−2)(1)(2)
− (−2)(1)(2)
− (3)(1)(−1)
− (−2)(−1)(−2)
+ (−2)(−1)(−1)
Therefore,
det(M ) = −6 − 4 + 4 + 3 + 4 − 2 = −1.

Example 3.2.2 Show that if A is a n×n upper triangular matrix, det(A) = a11 a22 a33 · · · ann .
Solution:
If δ = (1), δ(i) = i and aiδ(i) = aii . If δ ̸= (1), we have i > δ(i) for some 1 ≤ i ≤ n and,
in such case, aiδ(i) = 0. Hence,

a1δ(1) a2δ(2) a3δ(3) · · · anδ(n) = 0

except when δ = (1), where

a1δ(1) a2δ(2) a3δ(3) · · · anδ(n) = a11 a22 a33 · · · ann .

3.3 Some properties of determinants

1. Interchanging two rows changes the sign of the determinant

2. Adding one row to a multiple of another leaves the determinant unchanged

3. Multiplying a row by a non zero scalar α multiplies the determinant by α

Let X
det(A) = (signδ)a1δ(1) a2δ(2) · · · anδ(n)
δ∈Sn

and X
det(B) = (signδ)a1δ(1) a2δ(2) · · · λarδ(r) · · · anδ(n) .
δ∈Sn

Then,
det(B) = λdet(A).

18
 Consequently,
i. det(λA) = λn det(A)
ii. If one row in A is zero, det(A) = 0 since
X
det(A) = (signδ)a1δ(1) a2δ(2) · · · 0arδ(r) · · · anδ(n) = 0det(A) = 0.
δ∈Sn

a11 a12 a13 a11 0 0

4. 0 a22 a23 = 0 a22 0 = a11 a22 a33 , det(In ) =?
0 0 a33 0 0 a33

5. det(AB) = det(A)det(B)

6. det(AT ) = det(A)
7. Interchanging rows and columns of a determinant does not alter its value.
8. If two rows of A are identical, det(A) = 0.

3.4 Inverse of a Matrix

Consider matrix  
a a a
 11 12 13 
A =  a21 a22 a23 ,
 
 
a31 a32 a33
the cofactor matrix of A
 
A11 −A12 A13
 
C(A) =  −A21 A22 −A23
 

 
A31 −A32 A33

and adjoint of A denoted A∗ is A∗ = C(A)T .

A∗
Inverse of A, A−1 = |A|
If |A| = 0, then the inverse does not exist and A is said to be non invertible.

19
  
3 −1
Example 3.4.1 A =  , |A| = 8,
2 2
 
2 −2
C(A) =  ,
1 3
 
2 1
A∗ =  ,
−2 3
 
∗ ∗ 2 1
A A 1
A−1 = = =  
|A| 8 8 −2 3

Example 3.4.2  
8 −1 −3
 
B =  −5 1 ,
 
2
 
10 −1 −4
|B| = −1,  
−2 0 −5
 
C(B) =  −1 −2 −2  ,
 
 
1 −1 3
 
−2 −1 1
 
B ∗ =  0 −2 −1  ,
 
 
−5 −2 3
 
2 1 −1
∗ ∗
B B  
B −1 = = = 0 2 1
 
−1 |B| 


5 2 −3

20
3.4.1 Inverse by elementary operations
Theorem 3.4.3 If a square matrix A can be reduced to an identity matrix by a sequence
of elementary row operations, then it is invertible. Hence, its inverse can be obtained
by performing the same sequence of elementary row operations on the identity matrix
of the same dimension.

Perform elementary operations on the two sides until matrix A on the left is reduced to
an identity matrix. Whatever matrix the identity matrix on the right transforms into
is now the inverse of A.

A I
A−1 A A−1 I
I A−1

21
Chapter 4

SYSTEM OF LINEAR
EQUATIONS

4.1 System of Equations

4.1.1 Describing a System of equations

A system of n linear equations of the form

a11 x1 + a12 x2 + · · · + a1n xn = b1

a21 x1 + a22 x2 + · · · + a2n xn = b2
.. .. .. .
. . . = ..
an1 x1 + an2 x2 + · · · + ann xn = bn

can be transformed into

    
a11 a12 · · · a1n x1 b1
..
    
a21 · · · ···
    
 .  x2   b2 
=
.. .. .. .. .. ..
  
    
 . . . .  .   . 
    
an1 · · · · · · ann xn bn

22
 AX = B.
A is matrix of coefficient. If B = 0, the system is said to be homogeneous. Otherwise,
it is non-homogeneous.

4.1.2 Augmented matrix

Augmented matrix of the system above is
 
a a12 · · · a1n b1
 11
..

··· ···
 
 a21 . b2 
[A|B] = 
 .. .. .. .. .. 

 . . . . . 
 
an1 ··· · · · ann bn

Example 4.1.1
x1 + 6x2 + 2x3 = 1
−4x1 + x3 = 6
2x1 + 5x2 − 3x3 = 0
has     
1 6 2 x1 1
    
 −4 0 1   x2  =  6 
    
    
2 5 −3 x3 0
 
1 6 2 1
 
with augmented matrix  −4 0 1 6  is non-homogeneous.
 
 
2 5 −3 0

Example 4.1.2
5x1 − x2 + 2x3 = 0
x1 + 2x2 + 3x3 = 0
−x1 + 6x2 = 0

23
has     
5 −1 2 x 0
  1   
2 3   x2  =  0 
    
 1
    
−1 6 0 x3 0
 
5 −1 2 0
 
with augmented matrix  1 2 3 0  is homogeneous.
 
 
−1 6 0 0

4.2 Solution to System of Equations

4.2.1 Solving homogenous linear systems

1. For AX = 0, there is at least one trivial solution, which is X = 0. If any other
solution exists, it is non trivial.
2. It the row reduced form of A is In , the only solution is X = 0.
3. If Rank(A) is less than the number of linear equations, the system has other non
trivial solutions.

Example 4.2.1 The linear system

x1 + 2x3 = 0
2x1 − x2 + 3x3 = 0
4x1 + x2 + 8x3 = 0

has  
1 0 2
 
A =  2 −1 3
 

 
4 1 8
and its row reduced form  
1 0 0
 
AR =  0 1 0 .
 
 
0 0 1

24
    
x 0
 1   
Hence,  x2  =  0 
   
   
x3 0
Example 4.2.2 The linear system

x1 + 2x2 + 8x3 = 0
x1 − 3x2 − 7x3 = 0
−x1 + x2 + x3 = 0
has  
1 2 8
 
A =  1 −3 −7 
 
 
−1 1 1
and its row reduced form  
1 0 2
 
AR =  0 1 3 .
 
 
0 0 0
x1 + 2x3 = 0 ⇒ x1 = −2x3
x2 + 3x3 = 0 ⇒ x2 = −3x3
Hence,      
x −2x3 −2
 1     
 x2  =  −3x3  = λ  −3  ,
     
     
x3 x3 1
when x3 = λ.

4.2.2 Solving non-homogenous linear systems

For AX = B and B ̸= 0, let r and k respectively be the Rank(A) and the Rank([A|B])
and n the number of linear equations.
1. If r < k, the system in inconsistent and has no solutions
2. If r = k < n, the system has infinitely many solutions.
3. If r = k = n, the system has a unique solution.

25
Example 4.2.3 The linear system

x1 + 3x2 + x3 = 0
2x1 − x2 − x3 = 1
x1 − 4x2 − 2x3 = 2

has  
1 3 1 0
 
[A|B] =  2 −1 −1 1 
 
 
1 −4 −2 2
and its row reduced form
 
3
1 0 −2 7
 
[A|B]R =  0 1 73 − 17  .
 
 
1
0 0 0 −7

Hence, r = 2, k = 3 implies that the system is inconsistent.

Example 4.2.4 The linear system

x1 + 2x2 + 8x3 = 1
x1 − 3x2 − 7x3 = 6
−x1 + x2 + x3 = −4

has  
1 2 8 1
 
[A|B] =  1 −3 −7 6
 

 
−1 1 1 −4
and its row reduced form
 
1 0 2 3
 
[A|B]R =  0 1 3 −1  .
 
 
0 0 0 0

26
Hence, r = k = 2 < n = 3 implies that the system has infinitely many solutions.
x1 + 2x3 = 3 ⇒ x1 = −2x3 + 3
x2 + 3x3 = −1 ⇒ x2 = −3x3 − 1
Hence,        
x1 −2x3 + 3 −2x3 3
       
 x2  =  −3x3 − 1  =  −3x3  +  −1  .
       
       
x3 x3 x3 0
When x3 = λ.    
−2 3
   
X=λ  −3  +  −1 
   
   
1 0
| {z } | {z }
Gen. Sol. of AX=0 P articular sol. of AX=B

Example 4.2.5 Find the value(s) of α for the system

x+y−z = 2
x + 2y + z = 3
x + y + (α2 − 2)z = 2α
to have
(i) no solution
(ii) unique solution
(iii) infinitely many solution

Example 4.2.6 Find the value(s) of α for the system

x+y−z = 2
x + 2y + z = 3
x + y + (α2 − 5)z = α
to have
(i) no solution
(ii) unique solution
(iii) infinitely many solution

27
Example 4.2.7 Check whether or not the system
2x + 4y − 2z = 4
−x + 4y + 3z = 2
x − y + (α4 − 18)z = 2α

to have
(i) no solution
(ii) unique solution
(iii) infinitely many solution

28
Chapter 5

EIGENVALUES AND
EIGENVECTORS

5.1 Eigenvalues
Let A be × n matrix, I an identity matrix of the same dimension and x is a column
an n 
x1
 
 
 x2 
matrix 
 ..
. Then,

 . 
 
xn
Ax = λx (5.1.1)
which implies
(A − λI)x = 0 (5.1.2)
corresponds to the system of n homogeneous linear equations. It is easy to see that
x = 0 is a trivial solution of Eqn (5.1.2). However, if
n
X
CA (λ) = ai λi = |A − λI| = 0, (5.1.3)
i=0

then, Eqn (5.1.2) has other non-trivial solutions. This implies that if A − λI is singular,
Eqn (5.1.2) has non-trivial solution(s). The polynomial obtained from Eqn (5.1.3) is

29
called Characteristic Polynomial. If Eqn(5.1.3) is of degree n (that is A is of dimension
n × n), then there are n eigenvalues or characteristic roots λ1 , λ2 , λ3 , · · · λn .
 
3 1
Example 5.1.1 Find the eigenvalue of A =  .
2 2
 
λ 0
λIn =  .
0 λ

3−λ 1
|A − λI2 | = = (3 − λ)(2 − λ) − 2 = 0
2 2−λ
Solve (λ − 4)(λ − 1) = 0 ⇒ λ1 = 4, λ2 = 1.
 
3 −4
Example 5.1.2 Find the eigenvalue of A =  .
2 −6
 
λ 0
λIn =  .
0 λ

3−λ −4
|A − λI2 | = = (3 − λ)(−6 − λ) + 8 = 0
2 −6 − λ
Solve (λ − 2)(λ + 5) = 0 ⇒ λ1 = 2, λ2 = −5.
 
2 1 1
 
Example 5.1.3 Find the eigenvalue of A =  1 2 1 .
 
 
1 1 2
 
λ 0 0
 
λIn =  0 λ 0  .
 
 
0 0 λ

30
 2−λ 1 1
|A − λI3 | = 1 2−λ 1
1 1 2−λ
= (2 − λ)[(2 − λ)2 − 1] − 1[(2 − λ) − 1] + 1[1 − (2 − λ)] = 0
Solve (λ − 1)2 (λ − 4) = 0 ⇒ λ1 = 1, λ2 = 1, λ3 = 4.

5.2 Eigenvectors
To find eigenvectors, we solve for each case of λi the equation
(A − λi In )x = 0. (5.2.1)
The set of all these eigenvectors of A is called the eigenspace or row null space of A.
 
4 1
Example 5.2.1 Let A =  . λ1 = 1, λ2 = 5.
3 2
Case λ1
       
4−1 1 x 0 3 1 x
  1  =   =   1 
3 2−1 x2 0 3 1 x2
3x1 + x2 = 0 ⇒ x2 = −3x1 ,
     
x1 x1 1
 =  = x1   = ν1
x2 −3x1 −3
 
1
is the eigenvector corresponding to λ1 = 1. Simply  . Case λ2 = 5
−3
       
4−5 1 x 0 −1 1 x
  1  =   =   1 
3 2−5 x2 0 3 −3 x2
−x1 + x2 = 0 ⇒ x2 = x1 ,
       
x x 1 1
 1  =  1  = x 1   = x 2   = ν2
x2 x1 1 1

31
  
1
is the eigenvector corresponding to λ2 = 5. Simply  .
1
Example 5.2.2 Let  
2 0 1
 
A =  −1 4 −1  ,
 
 
−1 2 0
λ1 = 1, λ2 = 2, λ3 = 3.
Case λ1 = 1     
1 0 1 x 0
  1   
(A − λi In )X =  −1 3 −1   x2  =  0
    

    
−1 2 −1 x3 0
x1 + x3 = 0 ⇒ x1 = −x3 ,
− x1 + 3x2 − x3 = 0 ⇒ −x1 + 3x2 − (−x1 ) = 0 ⇒ x2 = 0.
       
x x −x3 −1
 1   1     
ν1 =  x2  =  0  =  0  = x3  0 
       
       
x3 x3 x3 1
or        
x x 1 1
 1   1     
ν1 =  x 2  =  0  = x1  0  =  0  .
       
       
x3 −x1 −1 −1
Case λ2 = 2     
0 0 1 x1 0
    
(A − λi In )X =  −1 2 −1   x2  =  0 
    
    
−1 2 −2 x3 0
x3 = 0, −x1 + 2x2 − x3 = 0 ⇒ −x1 + 2x2 = 0 ⇒ x1 = 2x2 .
     
x1 2x2 2
     
ν2 =  x 2  =  x 2  = x 2  1  .
     
     
x3 0 0

32
 Case λ3 = 3
    
−1 0 1 x 0
  1   
(A − λi In )X =  −1 1 −1   x2  =  0
    

    
−1 2 −3 x3 0

−x1 + x3 = 0 ⇒ x1 = x3 ,
− x1 + x2 − x3 = 0 ⇒ −x1 + x2 − x1 = 0 ⇒ −2x1 + x2 = 0 ⇒ 2x1 = x2 .
     
x x 1
 1   1   
ν3 =  x2  =  2x1  = x1  2  .
     
     
x3 x1 1

33
Chapter 6

CAYLEY-HAMILTON THEOREM
AND MINIMAL POLYNOMIAL

6.1 Cayley-Hamilton Theorem

Theorem 6.1.1 Let A be a non-singular square matrix and A∗ its adjoint. Then,
AA∗ = A∗ A = |A|I.

The following is Caley-Hamilton Theorem.

Theorem 6.1.2 Every square matrix is a root of its characteristic polynomial.

Proof 6.1.3 Recall that

n
X
CA (λ) = ai λi = |A − λI|.
i=0

Also, let
n−1
X
J(λ) = Ji λi ,
i=0

where J(λ) is the adjoint of A − λI (i.e. (A − λI)∗ ) and Ji are n × n matrices with
entries independent of λ. By Theorem 6.1.1,
(A − λI)J(λ) = |A − λI|I

34
so that
(A − λI)(Jn−1 λn−1 + · · · + J1 λ + J0 ) = (an λn + an−1 λn−1 + · · · + a1 λ + a0 )I.

By comparing coefficients of powers of λ, we obtain

−Jn−1 = an I
AJn−1 − Jn−2 = an−1 I
AJn−2 − Jn−3 = an−2 I (6.1.1)
.. .
. = ..
AJ1 − J0 = a1 I
AJo = a0 I
Multiplying each line (6.1.2) from the top respectively by An , An−1 , · · · , A, I, then
−An Jn−1 = A n an I
An Jn−1 − An−1 Jn−2 = an−1 An−1
An−1 Jn−2 − An−2 Jn−3 = an−2 An−2 (6.1.2)
.. .
. = ..
2
A J1 − AJ0 = a1 A
AJo = a0 I
Adding the RHS and the LHS,

RHS = LHS (6.1.3)

Xn
0 = ai Ai = CA (A) (6.1.4)
i=0
 
3 1
Example 6.1.4 A =  .
2 2

|A − λI2 | = (λ − 4)(λ − 1) = f (λ) = 0

    
−1 1 2 1 0 0
f (A) = (A − 4I)(A − I) =   = .
2 −2 2 1 0 0

35
  
2 1 1
 
Example 6.1.5 Let A =  1 2 1 .
 
 
1 1 2

2−λ 1 1
|A − λI3 | = 1 2−λ 1
1 1 2−λ

= (2 − λ)[(2 − λ)2 − 1] − 1[(2 − λ) − 1] + 1[1 − (2 − λ)] = 0

Check to see if (A − I3 )2 (A − 4I3 ) = 0.

Remark 6.1.6 Note that if A is a diagonal matrix with the diagonal entries d1 , d2 , · · · , dn ,
then
|A − λI| = (d1 − λ)(d2 − λ)(d3 − λ) · · · (dn − λ).

6.1.1 Minimal Polynomial

A minimal polynomial

m(x) = −α0 − α1 x − α2 x2 − · · · − αi−1 xi−1 + xi

is such that divides the characteristic polynomial CA (x). It can be obtained by finding
for which first i does the equation
n
X
Ai = αi−1 Ai−1 I
i=1

has solution, beginning with i = 1. That means first check

A = α0 I,

then
A2 = α0 I + α1 A,
then
A3 = α0 I + α1 A + α2 A2 ,
and so on. The first of these to equal zero is the minimal polynomial.

36
  
3 1
Example 6.1.7 Let A =  .
2 2

|A − λI2 | = (λ − 4)(λ − 1) = f (λ) = 0

Solving
A2 = α0 I + α1 A,
α0 = −4 and α1 = 5 so that the minimal polynomial is

−(−4+) − (5)x + x2 = 4 − 5x + x2 .
 
2 1 1
 
Example 6.1.8 Let A =  1 2 1 .
 
 
1 1 2

i. Find the minimal polynomial

ii. Use the minimal polynomial to compute A3 + 2A2 − A + 4I
iii. Find A−1
iv. Show that −8A−4 + 10A−3 + 2A−2 − 5A−1 + I = 0
Solution
i.
2−λ 1 1
|A − λI3 | = 1 2−λ 1
1 1 2−λ

= (2 − λ)[(2 − λ)2 − 1] − 1[(2 − λ) − 1] + 1[1 − (2 − λ)] = 0

Check to see if CA (A) = (A − I3 )2 (A − 4I3 ) = 0. Solving

A2 = α0 I + α1 A,
α0 = −4 and α1 = 5 so that the minimal polynomial is

−(−4+) − (5)x + x2 = 4 − 5x + x2 .

37
ii. Use m(x) = 4−5x+x2 to divide x3 +2x2 +4 and we have x +7


x2 − 5x + 4 x3 + 2x2 −x +4
− x3 + 5x2 − 4x

7x2 − 5x + 4
− 7x2 + 35x − 28

30x − 24
in which case,

x3 + 2x2 − x + 4 = (4 − 5x + x2 )(x + 7) + 30x − 24.

Hence, since m(A) = 4I − 5A + A2 = 0,

A3 + 2A2 − A + 4I = (4I − 5A + A2 )(A + 7I) + 30A − 24I = 30A − 24I.

Therefore,
 
36 30 30
 
A3 + 2A2 − A + 4I = 30A − 24I = 30 36 30 .
 
 
30 30 36

iii. Since 4I − 5A + A2 = 0, multiply through by A−1 so that

4I − 5A + A2 = 0
4A−1 − 5I + A = 0
4A−1 = 5I − A
1
A−1 = (5I − A)
4
Therefore,  
−1 −1
3
1 
A−1 = −1 3 −1
 
4 
−1 −1 3

iv. Multiply −8A−4 + 10A−3 + 2A−2 − 5A−1 + I = 0 by A4 and we obtain

−8I + 10A + 2A2 − 5A3 + A4 .

38
 Consider −8+10x+2x2 −5x3 +x4 and divide by m(x) to have x2 −


x2 − 5x + 4 x4 − 5x3 + 2x2 + 10x −
− x4 + 5x3 − 4x2

− 2x2 + 10x −
2x2 − 10x +

Therefore,
−8 + 10x + 2x2 − 5x3 + x4 = (4 − 5x + x2 )(x2 − 2).
Thus,
−8I + 10A + 2A2 − 5A3 + A4 = (4 − 5A + A2 )(A2 − 2).
Since
4 − 5A + A2 = 0,
−8I + 10A + 2A2 − 5A3 + A4 = 0.
Now multiply by A−4 and we have
−8A−4 + 10A−3 + 2A−2 − 5A−1 + I = 0.

Example 6.1.9 Find the minimum polynomial of matrix

 
2 8 0 0
 
 
0 2 0 0
C= 
.

0 0 4 2
 
0 0 1 3

Solution
Since C is a diagonal blocked matrix, the characteristic polynomial of C is the product
of the characteristic polynomials of its blocked diagonals. The diagonal matrices in C
are  
2 8
C1 =  
0 2
and  
4 2
C2 =  .
1 3

39
Their respective characteristic polynomials are

χ(C1 ) = (λ − 2)2
and
χ(C2 ) = (λ − 2)(λ − 5).
Hence, the characteristic polynomial of C, the product of χ(C1 ) and χ(C2 ) is

χ(C) = (λ − 2)3 (λ − 5).

The respective minimal polynomials of χ(C1 ) and χ(C2 ) are

m(C1 ) = (λ − 2)2
and
m(C2 ) = (λ − 2)(λ − 5).
Hence, minimal polynomial of C, the LCM of m(C1 ) and m(C2 ) is

m(C) = (λ − 2)2 (λ − 5).

Example 6.1.10 Find the minimum polynomial of matrix

 
2 1 0 0
 
 
0 2 0 0
C=
.

0 0 3 1
 
0 0 2 2

40
Chapter 7

SIMILAR, DIAGONAL AND

TRIANGULAR MATRICES

7.1 Similar Matrices

Definition 7.1.1 Two matrices A and B of the same dimension are said to be similar
if there is an invertible (or non singular) matrix P of the same dimension such that

B = P −1 AP.

Remark 7.1.2 Consider the eigenvectors associated with the eigenvalues of a matrix
A to be      
a b c
 1   1   1 
 a2  ,  b2  ,  c2 .
     
     
a3 b3 c3
The required matrix P can be obtained by arranging the eigenvectors of those eigenvalues
as the columns of P so that  
a b c
 1 1 1 
P =  a2 b 2 c 2 
 
 
a3 b 3 c 3

41
Example 7.1.3 Let  
8 −8 −2
 
A =  4 −3 −2  .
 
 
3 −4 1
The eigenvalues are λ1 = 1, λ2 = 2, λ3 = 3 and the eigenvectors are
     
4 3 2
     
 3 , 2 ,
     
1 
     
2 1 1

the matrix  
4 3 2
 
P =  3 2 1 ,
 
 
2 1 1
 
−1 1 1
 
P −1 =  1 0 −2 
 
 
1 −2 1
 
1 0 0
 
−1
and D = P AP =  0 2 0 
 
 
0 0 3

Remark 7.1.4 It can be observed that

 
λ 0 0
 1 
−1
D=P AP =  0 λ2 0
 

 
0 0 λ3
 
3 1
Example 7.1.5 Let A =  . The eigenvalues are λ1 = 1, λ2 = 4 and the
2 2

42
    
1 1
eigenvectors are  ,  the matrix
−2 1
   
1
1 1 − 13
P =  , P −1 =  3 
2 1
−2 1 3 3

and    
1 0 λ1 0
D = P −1 AP =  = .
0 4 0 λ2
 
4 1
Example 7.1.6 Let A =  . The eigenvalues are λ1 = 1, λ2 = 5 and the
3 2
  
1 1
eigenvectors are   ,   the matrix
−3 1
   
1
1 1 − 14
P =  , P −1 =  4 
3 1
−3 1 4 4

and    
1 0 λ1 0
D = P −1 AP =  = .
0 5 0 λ2

Remark 7.1.7 Note that similar matrices have the same determinant, characteristic
roots, rank and characteristic polynomials.

43
Chapter 8

SIMILAR AND TRIANGULAR

MATRICES

8.1 Similar Matrices

Definition 8.1.1 Two matrices A and B of the same dimension are said to be similar
if there is an invertible (or non singular) matrix P of the same dimension such that

B = P −1 AP.
     
a1 b1 c1
     
Remark 8.1.2 Consider the eigenvectors of a matrix to be  a2  ,  b2  ,  c 2 
     
     
a3 b3 c3
 
a a a
 1 2 3 
the matrix P =  b1 b2 b3 
 
 
c1 c2 c3
 
8 −8 −2
 
Example 8.1.3 Let A =  4 −3 −2 . The eigenvalues are λ1 = 1, λ2 = 2, λ3 = 3
 
 
3 −4 1

44
        
4 3 2 4 3 2
       
and the eigenvectors are  3  ,  2  ,  1  the matrix P =  3 2 1  and
       
       
2 1 1 2 1 1
 
1 0 0
 
D = P −1 AP =  0
 
2 0 
 
0 0 3
 
λ 0 0
 1 
Remark 8.1.4 It can be observed that D = P −1 AP =  0 λ2 0
 

 
0 0 λ3
 
3 1
Example 8.1.5 Let A =  . The eigenvalues are λ1 = 1, λ2 = 4 and the
2 2
       
1
1 1 1 1 3
− 31
eigenvectors are   ,   the matrix P =   , P −1 =  
2 1
−2 1 −2 1 3 3
   
1 0 λ 0
and D = P −1 AP =  = 1 .
0 4 0 λ2

Remark 8.1.6 Note that similar matrices have the same determinant, characteristic
roots, rank and characteristic polynomials.

8.2 Triangular Matrices

8.2.1 Upper and lower triangular matrices

   
a11 a12 a13 a11 0 0
   
U =  0 a22 a23  and L =  a21 a22 0  are upper and lower triangular
   
   
0 0 a33 a31 a32 a33
matrices respectively. Given that a11 ̸= 0, a22 ̸= 0, a33 ̸= 0, we can find triangular

45
matrices such that
A = L1 U1
or
A = U2 L2 .
 
2 −1 3
 
Example 8.2.1 Let A =  1 −1 . Decompose A into
 
2
 
1 1 1
    
2 0 0 a b c 2 −1 3
    
A = L1 U1 =  1   0 f d  =  1 2 −1 
    
g 0
    
1 1 1 0 0 e 1 1 1
and use this decomposition to solve the system
2x − y + 3z = 7 (8.2.1)
x + 2y − z = 1 (8.2.2)
x+y+z = 3 (8.2.3)
Solution:
Solving simultaneously the equations 2a = 2, 2b = −1,
2c = 3, b + 
f = 1, b + gf =

2 0 0 1 − 21 32
   
2, c + dg = −1, c + d + e = 1, we obtain L1 =  1 35 0  and U1 =  0 23 − 32
   

   
1 1 1 0 0 1
Note that
AX = Y ⇒ L1 U1 X = Y ⇒ U1 X = L−1 1 Y.
 
1
0 0
 2 
L−1
 3 3
=  − 10 5 0  .

1
 
− 15 − 35 1
Furthermore,
       
1 − 21 32 x 1
0 0 7 7
    2    2 
 0 23 − 32   y  =  − 10
3 3
0   1  =  − 23  .
        
    5    
1 3
0 0 1 z −5 −5 1 3 1
Hence, x = 2, y = 0, z = 1.

46
     
a b c 2 0 0 2 −1 3
    
Example 8.2.2 Also, A = U2 L2 =  0 f d   1 g 0  =  1 2 −1  .
    
    
0 0 e 1 1 1 1 1 1
 
5 8
− 3
 6 3 
Obtain and solve the appropriate simultaneous equations and we get U2 =  0 2 −1 
 
 
0 0 1
 
2 0 0
 
 3
and L2 =  1 2 0 

 
1 1 1

8.3 Reducing a System to Triangular Matrix

The system
2x − y + 3z = 7 (8.3.1)
x + 2y − z = 1 (8.3.2)
x+y+z = 3 (8.3.3)
can be represented as
    
2 −1 3 x 7
    
−1   y  =  1  .
    
 1 2
    
1 1 1 z 3
The augmented matrix is  
2 −1 3 7
 
 1 2 −1 1 .
 
 
1 1 1 3
When the matrix of coefficients is reduced to a triangular matrix we obtain
 
2 −1 3 7
 
 0 25 − 25 − 52  .
 
 
0 0 1 1

47
     
2 −1 3 x 7
    
 0 25 − 52   y  =  − 25 .
     
    
0 0 1 z 1
Hence, x = 2, y = 0, z = 1.

48
Chapter 9

VECTOR SPACES AND LINEAR

TRANSFORMATIONS

9.1 Vector Spaces

Definition 9.1.1 A vector space V over a field F is an additive abelian group such that
∀u, v ∈ V and scalars α, µ ∈ F
(i) αv ∈ V
(ii) (α + µ)v = αv + αv
(iii) α(u + v) = αu + αv
(iv) (αµ)v = α(µv) = µ(αv)
(v) ∃ 1 ∈ F such that 1v = v
Remark 9.1.2 Recall that R, Q and C are fields. Elements of a field are called scalar
and those of a vector space are called vectors.
Example 9.1.3 (i) The product of any field n times form a vector space such as Rn , Qn
and Cn . For instance,
Rn = R| ×R×R {z× · · · × R} .
n times
n
If x ∈ R , x = (x1 , x2 , · · · , xn ) and for λ ∈ R
λx = (λx1 , λx2 , · · · , λxn )
x + y = (x1 , x2 , · · · , xn ) + (y1 , y2 , · · · , yn ) = (x1 + y1 , x2 + y2 , · · · , xn + yn ) ∀x, y ∈ Rn .
The axioms of a vector space can be verified.

49
Example 9.1.4 (ii) The set of all n × m matrices with entries from a field F denoted
Mn,m (F) is also a vector space if we define addition and scalar multiplication respectively
as addition of matrices and scalar multiplication of matrices. That is for A = (aij ), B =
(bij ) ∈ Mn,m (F) and λ ∈ F,
A + B = (aij + bij ) and λA = (λaij ).
Example 9.1.5 (iii) Consider a non empty set X and a field F. Let F (X) be the set
of all functions of X with values in F. It is also a vector space if addition and scalar
multiplication are defined as follow: ∀ f (x), g(x) ∈ F (X) and λ ∈ F,
f (x) + g(x) = (f + g)(x)
λf (x) = (λf )(x),
where g0 (x) = 0 ∀x ∈ X is the zero function and for any f (x), there is a −f (x) in
F (X) such that −f (x) = (−f )(x) ∀x ∈ X.
Definition 9.1.6 Let v1 , v2 , · · · , vn be vectors in a vector space V over a field F. A
vector v ∈ V is said to be a linear combination of v1 , v2 , · · · , vn if for λ1 , λ2 , · · · , λn ∈ F,
n
X
v = λ1 v1 + λ2 v2 + · · · λn vn = λi vi .
i=1

Definition 9.1.7 The set v1 , v2 , · · · , vn in V is said to be linearly independent over F

if for λ1 , λ2 , · · · , λn ∈ F,
n
X
λi vi = 0 ⇒ λ1 = λ2 = · · · = λn = 0;
i=1

otherwise, it is said to be linearly dependent if

n
X
λi vi = 0
i=1

but not all λi ’s are zero.

Example 9.1.8 Let v = 3t2 + 5t − 5 = λ1 (t2 + 2t + 1) + λ2 (2t2 + 5t + 4) + λ3 (t2 + 3t + 6).
Find λ1 , λ2 , λ3 .
v = 3t2 + 5t − 5 = (λ1 + 2λ2 + λ3 )t2 + (2λ1 + 5λ2 + 3λ3 )t + (λ1 + 4λ2 + 6λ3 )
λ1 + 2λ2 + λ3 = 3
2λ1 + 5λ2 + 3λ3 = 5 .
λ1 + 4λ2 + 6λ3 = −5
We can find that λ1 = 3; λ2 = 1; λ3 = −2.

50
Example 9.1.9 Let (3, 7, −4) = λ1 (1, 2, 3) + λ2 (2, 3, 7) + λ3 (3, 5, 6). Find λ1 , λ2 , λ3 .
Definition 9.1.10 If every member of V can be expressed as a linear combination of the
set v1 , v2 , · · · , vn ∈ V, then the set is set to span or generate V. That is, v1 , v2 , · · · , vn ∈
V spans or generates V if λ1 , λ2 , · · · , λn ∈ F exist such that
n
X
v= λi vi ∀v ∈ V.
i=1

Remark 9.1.11 A vector space is said to be finitely generated if it is generated by a

finite set.
Xn
Definition 9.1.12 The collection of all linear combinations ( λi vi ) of the set of
i=1
vectors v1 , v2 , · · · , vn ∈ V is called span(v1 , v2 , · · · , vn ). It is also called the linear span
of v1 , v2 , · · · , vn .
Definition 9.1.13 The linearly independent subset of a vector space V over a field F
is called the basis for V if it spans V. That is, v1 , v2 , · · · , vn ∈ V is a basis if every
v ∈ V can be written as
n
X
v= λi vi = 0 ⇒ λ1 = λ2 = · · · = λn = 0;
i=1

Example 9.1.14 1. The standard basis for Fn is the set {ei = (0, 0, 0, · · · , 1, 0, · · · , 0)}
where the i-th entry is 1 and 0 elsewhere.
For instance, standard basis for R3 is e1 = (1, 0, 0), e2 = (0, 1, 0) and e3 = (0, 0, 1)
2. The basis for M2×2 (F) is
       
1 0 0 1 0 0 0 0
 , , , .
0 0 0 0 1 0 0 1

Remark 9.1.15 The dimension of a vector space is the number of elements in its basis.
Definition 9.1.16 A subset B of a vector space V over a field F is a subspace of V if
it is also a vector space over F under the same binary operations.
Definition 9.1.17 Let u, v ∈ B and α, β ∈ F. Then B is a subspace of V if and only if
αu + βv ∈ B.
Remark 9.1.18 The set span(v1 , v2 , · · · , vn ) = { ni=1 λi vi |λi ∈ F, vi ∈ V} is a sub-
P
space of V generated by {v1 , v2 , · · · , vn }. It is the intersection of all the subspaces of V
containing {v1 , v2 , · · · , vn }.

51
9.2 Some Results in Vector Space
Theorem 9.2.1 The non zero vectors {vi }ni=1 in a vector space V over a field F are
linearly dependent if one of the vectors is a linear combination of the others.

Proof 9.2.2 Suppose we have λ1 , λ2 , · · · , λn ∈ Fsuch that

vk = λ1 v1 + λ2 v2 + · · · + λk−1 vk−1 + λk+1 vk+1 + · · · + λn vn .
Then,

λ1 v1 + λ2 v2 + · · · + λk−1 vk−1 + (−1)vk + λk+1 vk+1 + · · · + λn vn = 0.

Since all λ’s are not zero, at least λk = −1, {vi }ni=1 are linearly dependent.

Lemma 9.2.3 Let {vi }ni=1 be non zero in a vector space V over a field F. They are
linearly dependent if and only if one of them is a linear combination of the preceding
ones.

Proposition 9.2.4 Any two basis for finitely generated vector space have the same
number of elements.

Proposition 9.2.5 Let U and W be subspaces of a finite dimensional vector space V

over F. Then,
(1) U + W = {u + w|u ∈ U, w ∈ W} is a subspace of V called linear sum of U and W.
(2) U ∩ W is a subspace of V
(3) dimU + dimW = dim(U + W).

9.3 Linear Mappings/Transformations

Definition 9.3.1 Let Vn (F) and Vm (F) be finite dimensional vector spaces. A mapping
T : Vn (F) → Vm (F) is called a linear transformation if for every u, v ∈ Vn (F) and λ ∈ F
(1) T (u + v) = T (u) + T (v)
(2) T (λu) = λT (u)

Definition 9.3.2 The set of all linear transformations from Vn (F) to Vm (F) is denoted
L(Vn (F), Vm (F)).

52
Remark 9.3.3 Let T1 , T2 ∈ L, then, T1 + T2 ∈ L. This is because for v ∈ Vn (F) and
λ∈F
(T1 + T2 )(v) = T1 (v) + T2 (v) ∈ L
(λ(T1 + T2 ))(v) = (λT1 + λT2 )(v) = λT1 (v) + λT2 (v) ∈ L
L can be seen as a set of homomorphisms of n-dimensional vector space Vn (F) to
m-dimensional space Vm (F) denoted HomF (Vn (F), Vm (F)). Any element of HomF is
called F-homomorphism.

Definition 9.3.4 Let T : Vn (F) → Vm (F). Then T

(i) is a monomorphism if it is one-to-one

(ii) is an epimorphism if it is onto
(iii) is an isomorphism if it is both onto and one-to-one
(iv) Kernel of T is Ker(T ) = {v ∈ Vn (F)|T (v) = 0}
(v) Image of T is Im(T ) = {T (v) ∈ Vm (F)|v ∈ Vn (F)}

Remark 9.3.5 Note that, for the linear transformation

T : Vn (F) → Vm (F),
(i) Kernel of T is a subspace of Vn (F)
(ii) Image of T is a subspace of Vm (F)

Definition 9.3.6 Given a linear transformation T : Vn (F) → Vm (F).

(i) Ker(T ) is called the null-space of T
(ii) Dimension of Ker(T ) is called the nullity of T
(iii) Dimension of Im(T ) is called the rank of T

Remark 9.3.7 Note that dim(Ker(T )) + dim(Im(T )) = dim(Vn (F))

Example 9.3.8 (i) The mapping T : R2 → R2 defined by T (x, y) = (x, 0) is a linear

transformation.
Consider (x1 , y1 ), (x2 , y2 ) ∈ R2 and α, β ∈ F.
T (α(x1 , y1 ) + β(x2 , y2 )) = T (αx1 + βx2 , αy1 + βy2 ) (9.3.1)
= (αx1 + βx2 , 0) (9.3.2)
= (αx1 , 0) + (βx2 , 0) (9.3.3)
= α(x1 , 0) + β(x2 , 0) (9.3.4)
= αT (x1 , y1 ) + βT (x2 , y2 ) (9.3.5)

53
(ii) Let V be a collection of functions of t. The derivative mapping D : V → V defined
by D(v) = dvdt
is a linear transformation.
Consider u, v ∈ V and α, β ∈ F.
d(αu + βv)
D(αu + βv) = (9.3.6)
dt
du dv
= α +β (9.3.7)
dt dt
= αD(u) + βD(v) (9.3.8)
(iii) Let VR be a collection of functions of t. The integral mapping I : V → R defined
1
by I(f (t)) = 0 f (t)dt is a linear transformation.
Consider u(t), v(t) ∈ V and α, β ∈ F.
Z 1
I(αu + βv) = (αu + βv)dt (9.3.9)
0
Z 1 Z 1
= αu dt + βv dt (9.3.10)
0 0
Z 1 Z 1
= α u dt + β v dt (9.3.11)
0 0
= αI(u) + βI(v) (9.3.12)

9.4 Change of Bases

Let c1 , c2 , · · · , cn be a basis for a vector space V of dimension n and f1 , f2 , · · · , fn
another basis such that
f1 = a11 c1 + a12 c2 + · · · + a1n cn
f2 = a21 c1 + a22 c2 + · · · + a2n cn
.. .. .. .. .. .. .. .. .. .
. = . . . . . . . .
fn = an1 c1 + an2 c2 + · · · + ann cn
Then,  
a11 a12 · · · a1n
..
 
a21 · · · ···
 
 . 
A=
 .. .. .. ..


 . . . . 
 
an1 · · · · · · ann

54
and
P = AT
is the transition or change of basis matrix from {ci } to {fi }. Furthermore, Q = P −1 is
the transition or change of basis matrix from {fi } to {ci }. If on the other hand,

c1 = b11 f1 + b12 f2 + · · · + b1n fn

c2 = b21 f1 + b22 f2 + · · · + b2n fn
.. .. .. .. .. .. .. .. .. .
. = . . . . . . . .
cn = bn1 f1 + bn2 f2 + · · · + bnn fn

Then,  
b11 b12 · · · b1n
..
 
b21 · · · ···
 
 . 
B=
 .. .. .. ..


 . . . . 
 
bn1 · · · · · · bnn
and
Q = P −1 = B T
is the transition or change of basis matrix from {fi } to {ci }.

Example 9.4.1 Consider that ei = {(1, 0), (0, 1)} and fi = {(2, 3), (1, 5)} are bases in
IR2 . Then, Find the transition matrix that changes the basis of V from
i. {ei } to {fi }
ii. {fi } to {ei }
Solution

i.
f1 = (2, 3) = α1 e1 + α2 e2
.
f2 = (1, 5) = β1 e1 + β2 e2
Hence,
(2, 3) = α1 (1, 0) + α2 (0, 1)
,
(1, 5) = β1 (1, 0) + β2 (0, 1)

55
 from where α1 = 2, α2 = 3, β1 = 1, β2 = 5. Thus,
 
2 3
A= 
1 5

and  
2 1
P = .
3 5

ii.  
5 −3
7 7
Q= 
−1 2
7 7

Example 9.4.2 Consider that ei = {(1, 0, 0), (0, 1, 0), (0, 0, 1)} and fi = {(3, 3, 4), (1, 2, 5), (2, 2, 2)}
are bases in IR3 . Then, Find the transition matrix that changes the basis of V from
i. {ei } to {fi }
ii. {fi } to {ei }
Solution

i.
(3, 3, 4) = a11 (1, 0, 0) + a12 (0, 1, 0) + a13 (0, 0, 1)
(1, 2, 5) = a21 (1, 0, 0) + a22 (0, 1, 0) + a23 (0, 0, 1) .
(2, 2, 2) = a31 (1, 0, 0) + a32 (0, 1, 0) + a33 (0, 0, 1)
Hence, a11 = a12 = 3, a13 = 4, a21 = 1, a23 = 5, a22 = a31 = a32 = a33 = 2 Thus,
 
3 3 4
 
A= 1
 
2 5 
 
2 2 2

and  
3 1 2
 
AT = P =  3 2 2
 

 
4 5 2

56
 ii.  
3 −4 1
 
Q =  −1 1
 
0 
 
−7 11 −3
2 2 2

Example 9.4.3 Consider that ei = {(1, 0), (0, 1)} and fi = {(1, 3), (2, 5)} are bases in
IR2 . Then, Find the transition matrix that changes the basis of V from
i. {ei } to {fi }
ii. {fi } to {ei }
Example 9.4.4 Consider that ci = {(1, 1, 0), (1, 2, 3), (1, 3, 5)} and fi = {(3, 3, 4), (−4, 7, 6), (−8, 10, 7)}
are bases in IR3 . Then, Find the transition matrix that changes the basis of V from
i. {ci } to {fi }
ii. {fi } to {ci }
Theorem 9.4.5 Let P be transition matrix that changes the basis of V from {ei } to
{fi } in a vector space V. Then, any linear operator T on V is such that
[T ]f = P −1 [T ]e P,
where P −1 transition matrix that changes the basis of V from {fi } to {ei } in the vector
space.
This theorem will be illustrated with an example.
Example 9.4.6 Consider that ei = {(1, 0, 0), (0, 1, 0), (0, 0, 1)} and fi = {(3, 3, 4), (1, 2, 5), (2, 2, 2)}
are bases in IR3 and T is the linear transformation
T : IR3 −→ IR3
defined by T (x, y, z) = (2x, z, y). Show that
[T ]f = P −1 [T ]e P,
where P −1 is transition matrix that changes the basis of V from {fi } to {ei } in the
vector space.
Solution

T (f1 ) = T (3, 3, 4) = (6, 4, 3) = a1 (3, 3, 4) + a2 (1, 2, 5) + a3 (2, 2, 2)

T (f2 ) = T (1, 2, 5) = (2, 5, 2) = b1 (3, 3, 4) + b2 (1, 2, 5) + b3 (2, 2, 2) .
T (f3 ) = T (2, 2, 2) = (4, 2, 2) = c1 (3, 3, 4) + c2 (1, 2, 5) + c3 (2, 2, 2)

57
From these, the matrix of coefficients is
 
5 −2 −7
2
 
 −12 3
 35 .

 2 
6 −2 −6

Hence, its transpose is the required

 
5 −12 6
 
[T ]f =  −2 −2  .
 
3
 
−7 35
2 2
−6

But note that

T (e1 ) = T (1, 0, 0) = (2, 0, 0) = f1 (1, 0, 0) + f2 (0, 1, 0) + f3 (0, 0, 1)

T (e2 ) = T (0, 1, 0) = (0, 0, 1) = g1 (1, 0, 0) + g2 (0, 1, 0) + g3 (0, 0, 1) .
T (e3 ) = T (0, 0, 1) = (0, 1, 0) = h1 (1, 0, 0) + h2 (0, 1, 0) + h3 (0, 0, 1)

The coefficient matrix from this is

 
2 0 0
 
.
 
 0 0 1
 
0 1 0

Its transpose is  
2 0 0
 
[T ]e =  0 0 1  .
 
 
0 1 0
Hence,
     
3 −4 1 2 0 0 3 1 2 5 −12 6
     
P −1 [T ]e P =  −1  =  −2 −2  = [T ]f .
     
1 0  0 0 1  3 2 2 3
     
−7 11 −3 −7 35
2 2 2
0 1 0 4 5 2 2 2
−6

58
Example 9.4.7 Consider that ei = {(1, 0), (0, 1)} and fi = {(1, 3), (2, 5)} are bases in
IR2 and T is the linear transformation

T : IR2 −→ IR2

defined by T (x, y) = (2x, 3x − y). Show that

[T ]f = P −1 [T ]e P,

where P −1 transition matrix that changes the basis of V from {fi } to {ei } in the vector
space.

Example 9.4.8 Consider that ei = {(1, 0, 0), (0, 1, 0), (0, 0, 1)} and fi = {(1, 1, 1), (1, 1, 0), (1, 0, 0)}
are bases in IR3 and T is the linear transformation

T : IR3 −→ IR3
defined by T (x, y, z) = (2y + z, x − 4y, 3x). Show that

[T ]f = P −1 [T ]e P,

where P −1 transition matrix that changes the basis of V from {fi } to {ei } in the vector
space.

59
Chapter 10

BILINEAR, QUADRATIC AND

CANONICAL FORMS

10.1 Bilinear Form

Definition 10.1.1 Let V be a vector space over a field IK. The mapping
f : V × V −→ IK
is said to be a bilinear form on V if
i. f (αu1 + βu2 , v) = αf (u1 , v) + βf (u2 , v)
ii. f (u, αv1 + βv2 ) = αf (u, v1 ) + βf (u, v2 ),
where α, β ∈ IK and u, u1 , u2 , v, v1 , v2 ∈ V.
Example 10.1.2 Let h and g be two linear functionals on a vector space V and f :
V × V −→ IK defined by f (u, v) = h(u)g(v). Show that f is a bilinear form on V.
Solution

f (αu1 + βu2 , v) = h[αu1 + βu2 ]g(v)

= [h(αu1 ) + h(βu2 )]g(v)
= [αh(u1 ) + βh(u2 )]g(v) .
= αh(u1 )g(v) + βh(u2 )g(v)
= αf (u1 , v) + βf (u2 , v)

60
Furthermore,
f (u, αv1 + βv2 ) = h(u)g(αv1 + βv2 )
= h(u)[g(αv1 ) + g(βv2 )]
= h(u)[αg(v1 ) + βg(v2 )] .
= αh(u)g(v1 ) + βh(u)g(v2 )
= αf (u, v1 ) + βf (u, v2 )
Hence, f is a bilinear form on V.
Proposition 10.1.3 Let f, g : V × V −→ IK be a bilinear form on a vector field V
over a field IK. Then, for any α, β, λ ∈ IK
i. f + g is also bilinear
ii. λf is also bilinear
Proof 10.1.4 i. We show that f + g is also bilinear.

(f + g)(αu1 + βu2 , v) = f (αu1 + βu2 , v) + g(αu1 + βu2 , v)

= αf (u1 , v) + βf (u2 , v) + αg(u1 , v) + βg(u2 , v)
.
= αf (u1 , v) + αg(u1 , v) + βf (u2 , v) + βg(u2 , v)
= α(f + g)(u1 , v) + β(f + g)(u2 , v)

Furthermore,

(f + g)(u, αv1 + βv2 ) = f (u, αv1 + βv2 ) + g(u, αv1 + βv2 )

= αf (u, v1 ) + βf (u, v2 ) + αg(u, v1 ) + βg(u, v2 )
.
= αf (u, v1 ) + αg(u, v1 ) + βf (u, v2 ) + βg(u, v2 )
= α(f + g)(u, v1 ) + β(f + g)(u, v2 )

ii. We now show that λf is also bilinear.

(λf )(αu1 + βu2 , v) = λf (αu1 + βu2 , v)

= λ[αf (u1 + βu2 , v)]
.
= λ[αf (u1 , v) + βf (u2 , v)]
= α(λf )(u1 , v) + β(λf )(u2 , v)

61
 Furthermore,

(λf )(u, αv1 + βv2 ) = λf (u, αv1 + βv2 )

= λ[αf (u, v1 ) + βf (u, v)2 ]
.
= αλf (u, v1 ) + βλf (u, v2 )
= α(λf )(u, v1 ) + β(λf )(u, v2 )

Example 10.1.5 For any n × n matrix A, show that the map

f (X, Y ) = X T AY,

where X and Y are column vectors with n entries, is a bilinear form on IK n .

Solution
Define
f (aX1 + bX2 , Y ) = (aX1 + bX2 )T AY
and
f (X, aY1 + bY2 ) = X T A(aY1 + bY2 ).
Hence,
f (aX1 + bX2 , Y ) = (aX1 + bX2 )T AY
= aX1T AY + bX2T AY .
= af (X1 , Y ) + bf (X2 , Y )
Similarly,
f (X, aY1 + bY2 ) = X T A(aY1 + bY2 )
= (X T A)(aY1 ) + (X T A)(bY2 )
.
= aX T AY1 + bX T AY2
= af (X, Y1 ) + bf (X, Y2 )

Remark 10.1.6 A polynomial

n X
X n
f (x, y) = aij xi yj (10.1.1)
i=1 j=1

62
which can be written as a matrix form
  
a11 a12 · · · a1n y1
..
  
 a21 · · · ···
 
  .  y2 
x1 x2 · · · xn 
 .. .. .. ..

 ..


 . . . .  . 
  
an1 · · · · · · ann n

is called a bilinear polynomial.

Example 10.1.7 Write the bilinear polynomial
f (x, y) = 3x1 y1 − 4x1 y2 + 2x2 y1 + 6x1 y3 + 2x3 y1 − x2 y2 − 5x3 y2 − x2 y3 − x3 y3
in a matrix form.
  
3 −4 6 y
   1 
x3  2 −1 −1   y2  .
  
x1 x2
  
2 −5 −1 y3

Example 10.1.8 Write the bilinear polynomial

f (x, y) = 3x1 y1 − 2x1 y2 + 5x2 y1 + 6x1 y3 + 2x3 y1 + 7x2 y2 + 4x3 y2 − 8x2 y3 − x3 y3
in a matrix form.
Example 10.1.9 Is the polynomial
f (x, y) = 2x1 y2 − 3x2 y1
in a matrix form?
Definition 10.1.10 Let f be a bilinear form on V and B = {c1 , c2 , · · · , cn } the basis
for V. The matrix A of the form
 
f (c1 , c1 ) f (c1 , c2 ) · · · f (cn , cn )
..
 
··· ···
 
 f (c2 , c1 ) . 
A= 
.. .. .. ..


 . . . . 
 
f (cn , c1 ) ··· · · · f (cn , cn )

is called the matrix representation of f relative to B.

63
Example 10.1.11 Define the bilinear form on IR2 by
f [(x1 , x2 ), (y1 , y2 ) = 2x1 y1 − 3x1 y2 + x2 y2 .

Find the matrix A in the basis

i. B = {u1 = (1, 0), u2 = (1, 1)}
ii. B = {v1 = (0, 1), v2 = (−1, 0)}
Solution
i. In A, a11 = f [u1 , u1 ] = f [(1, 0), (1, 0)] = 2, a12 = f [u1 , u2 ] = f [(1, 0), (1, 1)] = −1,
a21 = f [u2 , u1 ] = f [(1,1), (1, 0)] = 2 and a22 = f [u2 , u2 ] = f [(1, 1), (1, 1)] = 0.
2 −1
Hence, A =  .
2 0

ii. In A, a11 = f [v1 , v1 ] = f [(0, 1), (0, 1)] = 1, a12 = f [v1 , v2 ] = f [(0, 1), (−1, 0)] = 0,
a21 = f [v2 , v1 ] =
 f [(−1,0), (0, 1)] = 3 and a22 = f [v2 , v2 ] = f [(−1, 0), (−1, 0)] =
1 0
2. Hence, A =  .
3 2

10.2 Orthogonal Diagonalization of a Real Symmet-

ric Matrix
Recall that the norm of a vector x ∈ IRn is
v
u n
uX
||x|| = t x2i .
i=1

Hence, the normalized form of vector (1, 3, 2) is

1 3 2
( √ , √ , √ ).
14 14 14
Definition 10.2.1 Consider a square matrix A. It is said to be
i. symmetric if AT = A
ii. orthogonal if AAT = I = AT A, that is AT = A−1

64
 iii. is orthonormal if it is orthogonal and each of its column is normalized or of length
1
Definition 10.2.2 Consider a square matrix A. It is said to be orthogonally diago-
nalizable if we can find an orthogonal matrix P such that A is similar to a diagonal
matrix. That is,
D = P −1 AP = P T AP
Remark 10.2.3 If we have a matrix P̄ whose columns are the eigenvectors corre-
sponding to the eigenvalues of the symmetric matrix A, we can obtain an orthogonal
(orthonormal) matrix P whose columns are the normalized columns of P̄ , in which case,
P will diagonalize A.
 
3 −6 0
 
Example 10.2.4 Let A =  −6 0 6 . Obtain an orthogonal matrix P which
 
 
0 6 −3
diagonalizes A.
Solution
Obviously, A is symmetric. Also,
 
1
1 − 2 −2
 
 1
P̄ =  2 −1 2 

 
1 1 1

and  
2 −1 −2
1 
P =  1 −2 2  .
 
3 
2 2 1
Hence,  
0 0 0
 
D = P T AP =  0 −9 0 .
 
 
0 0 9
 
1 −2
Example 10.2.5 Let A =  . Obtain an orthogonal matrix P which diag-
−2 1
onalizes A.

65
Solution
Obviously, A is symmetric. Also,
 
1 −1
P̄ =  
1 1

and  
1 1 −1
P =√  .
2 1 1
Hence,  
−1 0
D = P T AP =  .
0 3

10.3 Quadratic Forms

Consider an n × n symmetric matrix A and a column matrix
 
x
 1 
 
 x2 
x= . 

.
 .. 
 
xn

A quadratic form is the expression

xT Ax = a11 x21 + a22 x22 + a33 x23 + 2a12 x1 x2 + 2a13 x1 x3 + 2a23 x2 x3 . (10.3.1)

To simply put,
n X
X n
xT Ax = aij xi xj . (10.3.2)
i=1 j=1

66
10.4 Canonical Forms
When we let x = P y, where  
y1
 
 
 y2 
y=
 ..


 . 
 
yn
and P is the orthogonal matrix that diagonalizes A, then,

xT Ax = (P y)T A(P y) = y T (P T AP )y.

This new expression y T (P T AP )y reduces xT Ax to a quadratic diagonal of the form

y T (P T AP )y = d11 y12 + d22 y22 + d33 y32 (10.4.1)

and removes the cross products. The expression in Eqn (10.4.1) is the canonical form
of Eqn (10.3.1).

Example 10.4.1 Consider the matrix

 
2 0 −2
 
A= 0 4 0 .
 
 
−2 0 5

i. Find the real quadratic form in variables x1 , x2 , x3 .

ii. Find the canonical form of the real quadratic form
Solution

i.
xT Ax = 2x21 + 4x22 + 5x23 − 4x1 x3 .
ii.  
√2 0 √1
 5 5 
P = 0
 
1 0 
 
√1 0 − √25
5

67
and  
1 0 0
 
P T AP =  0 4 0  .
 
 
0 0 6
Hence,
y T (P T AP )y = y12 + 4y22 + 6y32 .

68
Bibliography

[1] S.A. Ilori and O. Akinyele, Elementary Abstract and Linear Algebra, Univeristy
Press, Ibadan (1986).

69