CHAPTER 1

SYSTEMS OF LINEAR EQUATIONS

AND MATRICES

1
 CONTENTS
1.1 Introduction to Systems of Linear Equations
1.2 Gaussian Elimination
1.3 Matrices and Matrix Operations
1.4 Inverses; Rules of Matrix Arithmetic
1.5 Finding Inverse by Elementary Matrices
1.6 Further Results on Systems of Equations
and Invertibility
1.7 Diagonal, Triangular & Symmetric Matrices
2
1.1. INTRODUCTION TO SYSTEMS
OF LINEAR EQUATIONS

1.1.1. Linear Equations

1.1.2. Linear Systems
1.1.3. Augmented Matrices
1.1.4. Elementary Row Operations

3
 1.1.1 LINEAR EQUATIONS
Any straight line in the xy-plane is represented by a
linear equation of the form
a1x + a2y = b
where a1, a2, and b are real constants and a1, a2 are
not simultaneously zeros

More generally, a linear equation in the variables

(unknowns) x1, x2, …, xn has the form
a1x1 + a2x2 + … + anxn = b
where a1, a2, …, an and b are real constants
4
Example. The equations
x + 3 y = 7 y = 12 x + 3z + 1 x1 − 2 x2 − 3x3 + x4 = 7
are linear, while the equations
x + 3 y = 5 3x + 2 y − z + xz = 4 y = sin x
are not linear

A solution of a linear equation

a1x1 + a2x2 + … + anxn = b
is an n-tuple of real numbers s1, s2, …, sn such that the
equation is satisfied when we substitute x1= s1, x2= s2, …,
xn = sn.
The set of all solutions is called the solution set 5
Example. Finding the solution set of
(a) 4x – 2y = 1 (b) x1 – 4x2 + 7x3 = 5

Solution. (a) We assign to x an arbitrary value t

x=t
4t – 2y = 1
2y = 4t – 1
y = 2t − 12
The arbitrary value t is called a parameter. Substituting
t by any specific value will give a particular solution.
For example with t = 3 we get

x = 3 and y = 112
6
We can also assign to y an arbitrary value t, then
x = 12 t + 14 y =t
Note. The two sets of solutions are the same.

For example with t = 3 in the first set we have

x = 3 y = 112
this solution corresponds to the solution in the second
set with t = 11
2

Solution. (b) We assign arbitrary values to any two

variables and solve for the third one. For example
x1 = 5 + 4s – 7t x2 = s x3 = t
7
 1.1.2 LINEAR SYSTEMS

A system of linear equations (or a linear system ) is a

finite set of linear equations in the variables x1, x2, …, xn

A solution of a linear system is an n-tuple of real

numbers s1, s2, …, sn such that each equation is
satisfied when we substitute x1= s1, x2= s2, …, xn = sn.

8
Example. Consider the system
4x1 – x2 + 3x3 = –1
3x1 + x2 + 9x3 = –4
Then x1 = 1, x2 = 2, x3 = –1 is a solution.
However, x1 = 1, x2 = 8, x3 = 1 is not a solution since
these values satisfy only the first equation.

Note. Not all systems of linear equations have

solutions. For example the system
x + y=4
2x + 2y = 6
has no solution. We say that it is inconsistent
A system with at least one solution is called consistent
9
A general system of two linear equations in two
unknowns has the form:
a1x + b1y = c1
a2x + b2y = c2
The graph of these equations is two lines l1 and l2.
A pair x, y is a solution of the system if the point with
coordinates (x,y) is a point of intersection of l1 and l2.
There are 3 cases

✓The lines l1 and l2 are

parallel: there is no
intersection and the system
is inconsistent
10
 ✓The lines l1 and l2
intersect at only one point:
The system has a unique
solution

✓The lines l1 and l2 coincide:

The system has infinitely
many solutions
In general

A system of linear equations may have no solution,

exactly one solution or infinitely many solutions
11
An arbitrary system of m linear equations in n
unknowns has the form:
a11 x1 + a12 x2 + + a1n xn = b1
a21 x1 + a22 x2 + + a2 n xn = b2

am1 x1 + am 2 x2 + + amn xn = bm
The double subscripting on the coefficients of the
unknowns allows us to specify their location:
- The first subscript on the coefficient aij indicates
the equation in which this coefficient occurs (i)
- The second subscript indicates which unknown this
coefficient multiplies (j)
12
 1.1.3 AUGMENTED MATRICES

The general system of m linear equations in n unknowns

may be abbreviated by using the two-dimensional array

 a11 a12 a1n b1 

a a 
a2 n b2 
 21 22
 
 
 am1 am 2 amn bm 

which is called the augmented matrix for the system

13
Example. The augmented matrix for the system
x1 + x2 + 2 x3 = 9
2 x1 + 4 x2 − 3x3 = 1
3x1 + 6 x2 − 5 x3 = 0
is
1 1 2 9 
2 4 − 3 1 
 
3 6 − 5 0

Note. The unknowns must be written in the same

order, and the constant terms must be on the right
14
 The basic method for solving a system of linear
equations is to replace it by a new system with the
same solution set (the equivalent system) but which is
easier to be solved.
This is done in a series of steps by

1. Multiplying an equation through by a nonzero constant

2. Interchanging two equations
3. Adding a multiple of one equation to another

This results in a series of operations on the augmented

matrix called elementary row operations
15
 Elementary row operations
1. Multiply a row through by a nonzero constant
2. Interchange two rows
3. Add a multiple of one row to another

Example. Solve the following system by carrying out

elementary row operations on the augmented matrix

x + y + 2z = 9 1 1 2 9
 
2 x + 4 y − 3z = 1  
aug :=  2 4 -3 1
 
3x + 6 y − 5 z = 0  
3 6 -5 0
Solution. Add –2 times the 1st equation to the 2nd
16
Add –2 times the 1st equation to the 2nd
x + y + 2z = 9 1 1 2 9
 
 
2 y − 7 z = −17 aug :=  0 2 -7 -17
 
3x + 6 y − 5 z = 0  
3 6 -5 0
Add –3 times the 1st equation to the 3rd
x + y + 2z = 9 1 1 2 9
 
 
2 y − 7 z = −17 aug := 0 2 -7 -17
 
 
3 y − 11z = −27 0 3 -11 -27 
1
Multiply 2nd equation by 2
 1 1 2 9 
 
x + y + 2z = 9 
 -7

-17 
aug :=  0 1 
y − 72 z = − 172 
 2 2 
 
3 y − 11z = −27  0 3 -11 -27 
17
 Add –3 times the 2nd equation to the 3rd
 1 
x + y + 2z = 9
1 2 9
 
 
 -7 -17 
 
y − 72 z = − 172 aug := 
0 1
2 2 
 
 -3 
− 12 z = − 32  -1
 0 0 
 2 2 
Multiply 3rd equation by –2  1 1 2 9 
x + y + 2z = 9  
 
 -7 -17 
aug :=  0 
y− z =−
7
2
17
2


1
2 2 
 
z= 3  0 0 1 3 
Add –1 times the 2nd equation to the 1st
 11 35 
 1 0 
 2 
x + z=
11
2
35
2 

2

aug :=  -17 
y − 72 z = − 172
-7
 0 1 
 2 2 
 
z= 3 
 0 0 1

3 18
  11 35 
x + z=
11 35

 1 0 
2 2
 2 2 
y − 72 z = − 172  
aug :=  -7 -17 
 0 1 
z= 3 

2 2 

 
 0 0 1 3 
Add − 112 times the 3rd equation to the 1st , and 7
2
times the 3rd equation to the 2nd
1 0 0 1
x =1  
 
y =2 aug :=  0 1 0 2
 
 
z =3 0 0 1 3

The unique solution x = 1, y = 2, z = 3 is now clear

19
1.2. GAUSSIAN ELIMINATION

1.2.1. Echelon Forms

1.2.2. Elimination Methods
1.2.3. Back Substitution
1.2.4. Homogeneous Linear Systems

20
 1.2.1 ECHELON FORMS
In the last section a linear system in the unknowns x, y, z
is solved by reducing the augmented matrix to the form
1 0 0 1
 
 
aug :=  0 1 0 2
 
 
0 0 1 3

from which the solution x =1, y =2, z =3 becomes clear

This is an example of a matrix in reduced row-echelon

form
21
A row is said to be zero row if all of its entries are zero.
Otherwise it is said to be a nonzero row.
The first nonzero element in a nonzero row is called a
leading entry.

Definition. A matrix is said to be in row-echelon

form if it satisfies the following two properties:
1. The zero rows are grouped together at the bottom,
2. The leading entry of each nonzero row occurs
further to the right than the leading entry in the row
above it.

22
 A row-echelon matrix has the form of a staircase
     0       
 0    0
   0 0     
 0 0   0 0 0 0    
   
 0 0 0 0 0 0 0 0 0 0  
Where the entry  may be any nonzero number (the
leading entry), and  may be any number

Definition. A row-echelon matrix is said to be in

reduced row-echelon form if it also satisfies:
3. The leading entry in each nonzero row is 1 (leading 1)
4. Each leading 1 is the only nonzero entry in its column.
23
Example. The following matrices are in reduced row-
echelon form
 0 1 -2 0 1
1 4  1 0  

0 0
 
0
    0 0
     0 0 0 1 3  
0 7  0 0    

1 0
 
1
    0 0
     0 0 0 0 0
0 -1  0 1  
0 1 0  
 0 0 0 0 0

The following matrices are in row-echelon form but

not in reduced row-echelon form

1 4 -3 7  1 1 0 0 1 2 6 0
     
     
0 1 6 2  0 1 0 0 0 1 -1 0
     
     
0 0 1 5  0 0 0 0 0 0 0 1

24
Example. Solve the following systems where the
augmented matrices are in reduced row-echelon form.
1 0 0 5
 
 
(a) 0 1 0 -2
 
 
0 0 1 4

Solution. The corresponding system of equations is

x1 = 5
x2 = −2
x3 = 4
From this we can read out the solution immediately
x1 = 5, x2 = –2 , x3 = 4

25
 1 0 0 4 -1
 
(b)  
0 1 0 2 6
 
 
0 0 1 3 2

Solution. The corresponding system of equations is

x1 + 4 x4 = −1
x2 + 2 x4 = 6
x3 + 3 x4 = 2

Since x1, x2, x3 correspond to the leading entries in the

augmented matrix, we call them the leading variables,
the other variables (x4) are free variables
26
Solving for the leading variables in terms of the free
variables we obtain
x1 = −1 − 4 x4
x2 = 6 − 2 x4
x3 = 2 − 3x4

We can assign an arbitrary value, say t to the free

variable x4 and get the corresponding values for the
leading variables x1, x2, x3
x1 = –1 – 4t , x2 = 6 – 2t , x3 = 2 – 3t , x4 = t
There are infinitely many solutions.

27
  1 6 0 0 4 -2
 
(c)  
 0 0 1 0 3 1
 
 
 0 0 0 1 5 2
 
 
 0 0 0 0 0 0
Solution. The last row doesn't contribute to find the
solutions of the system. We can write the system as
x1 + 6 x2 + 4 x5 = −2
x3 + 3x5 = 1
x4 + 5 x5 = 2
We can solve for the leading variables x1, x3, x4 in
terms of the free variables x2 and x5
x1 = −2 − 6 x2 − 4 x5
x3 = 1 − 3x5
x4 = 2 − 5 x5 28
We can assign arbitrary values, say s and t to the free
variable x2 and x5. The general solution is

x1 = –2 – 6s – 4t , x2 = s, x3 = 1 – 3t ,
x4 = 2 – 5t, x5 = t

1 0 0 0
 
 
(d) 0 1 2 0
 
 
0 0 0 1

Solution. The last row corresponds to the equation

0x1 + 0x2 + 0x3 = 1
which has no solution. So the system has no solution
29
 1.2.2 ELIMINATION METHODS
We have seen that a linear system is easily solved if its
augmented matrix is in reduced row-echelon form

The Gaussian Elimination procedure (also called

row-reduction) allows us to put a matrix in a row-
echelon form. It consists of five steps.
We will present these steps via an example

0 0 -2 0 7 12
 
 
2 4 -10 6 12 28
 
 
2 4 -5 6 -5 -1 30
Step 1. Locate the leftmost nonzero column: column 1
0 0 -2 0 7 12
 
 
2 4 -10 6 12 28 
 
 
2 4 -5 6 -5 -1 

Step 2. Interchange the 1st row with another row, if

necessary, (row 2) so that the top of the column found
in Step 1 is nonzero

2 4 -10 6 12 28
 
 
0 0 -2 0 7 12 
 
 
2 4 -5 6 -5 -1 
31
Step 3. Multiply the 1st row with a suitable scalar (1/2)
to introduce a leading 1
1 2 -5 3 6 14
 
 
0 0 -2 0 7 12
 
 
2 4 -5 6 -5 -1

Step 4. Add a suitable multiple of the 1st row (-2) to

the rows below (row 3) so that all entries below the
leading 1 becomes zeros
1 2 -5 3 6 14
 
 
0 0 -2 0 7 12 
 
 
0 0 5 0 -17 -29 
32
Step 5. Cover the 1st row and begin again with Step 1
applied to the remaining submatrix. Continue Step 5
until the matrix is in row-echelon form
1 2 -5 3 6 14
Multiply new row 1 by – 



0 0 -2 0 7 12
1/2 to introduce leading 1  
 
0 0 5 0 -17 -29

 1 2 -5 3 6 14
 
 
Add –5 times new row 1 
 0 0 1 0
-7 
-6
 2 
to new row 2  
 
 0 0 5 0 -17 -29

 1 2 -5 3 6 14
Cover the new row 1 and 



 -7 
begin again with Step1 
 0 0 1 0
2
-6
 
 
 1 
 1
 0 0 0 0
 2 33 
  1 2 -5 3 6 14
 
Multiply new row 1 by 2 
 -7


 0 0 1 0 -6
to introduce leading 1 
 2 
 
 1 
 1
 0 0 0 0
 2 

 1 2 -5 3 6 14  1 2 -5 3 6 14
   
   
 -7   -7 
 0 0 1 0 -6  0 0 1 0 -6
 2   2 
   
   
 0 0 0 0 1 2  0 0 0 0 1 2

The matrix is now in row-echelon form.

34
 Note. Gaussian Elimination procedure as presented
is convenient for hand computations
Traditionally, Step 3 is not included since it causes
more computations

Step 3. Multiply the 1st row with a suitable scalar to

introduce a leading 1

Without this step the result is slightly different.

A call of MAPLE's command gausselim produces

 2 4 -10 6 12 28
 
 
 0 0 -2 0 7 12
 
 
 1 
 0 0 0 0 1
 2  35
 2 4 -10 6 12 28  1 2 -5 3 6 14
   
   
 0 0 -2 0 7 12  -7 
   0 0 1 0 -6
   2 
 1   
 0 0 0 0 1  
 2   0 0 0 0 1 2

Compare this with the result from Gaussian Elimination

procedure in which Step 3 is included
Note. The position of the leading entries is the same. They
are called pivot positions.
The nonzero number at a pivot position is called a pivot

The Gauss-Jordan Elimination procedure allows to find

the reduced row-echelon form. It includes another Step

36
 Step 6. Working upward from the last nonzero row. Add
suitable multiples of each row to the rows above it so
that all entries above the pivot becomes zeros
 1 2 -5 3 6 14
1 2 -5 3 6 14
  (2)+7/2*(3)
   
 -7   
 0

0 1 0 -6

0 0 1 0 0 1
 2   
   
 0 0 0 0 1 2 0 0 0 0 1 2

1 2 -5 3 0 2 1 2 0 3 0 7
   
   
0 0 1 0 0 1 0 0 1 0 0 1
   
  (1) + 5*(2)  
0 0 0 0 1 2 0 0 0 0 1 2

The matrix is now in reduced row-echelon form

37
Example. Solve by Gauss-Jordan elimination.
x1 + 3x2 − 2 x3 + 2 x5 = 0
2 x1 + 6 x2 − 5 x3 − 2 x4 + 4 x5 − 3x6 = −1
5 x3 + 10 x4 + 15 x6 = 5
2 x1 + 6 x2 + 8 x4 + 4 x5 + 18 x6 = 6

Solution. The augmented matrix of the system is

1 3 -2 0 2 0 0
 
 
2 6 -5 -2 4 -3 -1
 
 
0 0 5 10 0 15 5
 
 
2 6 0 8 4 18 6

38
Apply the row-reduction to the augmented matrix
1 3 -2 0 2 0 0 1 3 -2 0 2 0 0
   
   
0 0 -1 -2 0 -3 -1 0 0 -1 -2 0 -3 -1
   
   
0 0 5 10 0 15 5 0 0 0 0 0 0 0
   
   
0 0 4 8 0 18 6 0 0 0 0 0 6 2

1 3 -2 0 2 0 0
 
 
0 0 -1 -2 0 -3 -1
 
 
0 0 0 0 0 6 2
 
 
0 0 0 0 0 0 0

The last matrix is now in row-echelon form.

Apply step 6 to put it in reduced row-echelon form
39
Apply step 6 to put it in reduced row-echelon form
1 3 -2 0 2 0 0  1 3 -2 0 2 0 0
   
   
 0 0 -1 -2 0 0 0
0 0 -1 -2 0 -3 -1  
   
   1
0 0 0 0 0 6 2  0 0 0 0 0 1 
   3 
  
0 0 0 0 0 0 0  
 0 0 0 0 0 0 0

 1 3 0 4 2 0 0
 
 
 0 0 1 2 0 0 0
 
 
 1
 0 0 0 0 0 1 
 3 

 
 0 0 0 0 0 0 0

The last matrix is now in reduced row-echelon form

40
The corresponding system of equations is.
x1 + 3x2 + 4 x4 + 2 x5 = 0
x3 + 2 x4 = 0
x6 = 13

Solving for the leading variables

x1 = − 3x2 − 4 x4 − 2 x5
x3 = − 2 x4
x6 = 13
assigning arbitrary values to the free variables
x1 = −3r − 4s − 2t x2 = r x3 = −2s
x4 = s x5 = t x6 = 13
41
 1.2.3 BACK-SUBSITUTION
In the last example, we can stop at the row-echelon
form with Gaussian elimination
1 3 -2 0 2 0 0
 
 
0 0 -1 -2 0 -3 -1
 
 
0 0 0 0 0 6 2
 
 
0 0 0 0 0 0 0
The corresponding system is
x1 + 3x2 − 2 x3 + 2 x5 = 0
− x3 − 2 x4 − 3x6 = − 1
6 x6 = 2
42
 x1 + 3x2 − 2 x3 + 2 x5 = 0
− x3 − 2 x4 − 3x6 = − 1
6 x6 = 2

Solving backward (upward) for the leading variables

and assigning arbitrary values to the free variables

x6 = 13 x5 = t x4 = s
x3 = 1 − 2 x4 − 3x6 = 1 − 2s − 3( 13 ) = −2s x2 = r
x1 = −3x2 + 2 x3 − 2 x5 = −3r + 2(−2s) − 2t = −3r − 4s − 2t

This is the same solution as before.

The arbitrary values r, s, t are the parameters
43
The Gauss-Jordan elimination is convenient for hand
computations

However, when programmed to run on computers,

Gaussian elimination is faster than Gauss-Jordan
elimination (about 50%)

44
 1.2.4 HOMOGENEOUS LINEAR
SYSTEMS
A system of linear equation is said to be homogeneous
if all the constant terms are zero
a11 x1 + a12 x2 + + a1n xn = 0
a21 x1 + a22 x2 + + a2 n xn = 0

am1 x1 + am 2 x2 + + amn xn = 0
x1= 0, x2= 0, …, xn = 0 is always a solution called the
trivial solution.
All other solutions, if exist, are non trivial
45
Thus a homogeneous linear system either has
✓ only the trivial solution, or
✓ infinitely many solutions
A general homogeneous linear system of two
equations in two unknowns has the form:
a1x + b1y = 0
a2x + b2y = 0

The graphs of the two equations are two lines

passing through the origin: the trivial solution
corresponds to the origin (0, 0)

46
Example. Solve by Gauss-Jordan elimination.
2 x1 + 2 x2 − x3 + x5 = 0
− x1 − x2 + 2 x3 − 3x4 + x5 = 0
x1 + x2 − 2 x3 − x5 = 0
x3 + x4 + x5 = 0
Solution. The augmented matrix of the system
and its reduced row-echelon form are
 2 2 -1 0 1 0  1 1 0 0 1 0
   
   
 -1 -1 2 -3 1 0  0 0 1 0 1 0
   
   
 1 1 -2 0 -1 0  0 0 0 1 0 0
   
   
 0 0 1 1 1 0  0 0 0 0 0 0
47
The corresponding system of equations is
x1 + x2 + x5 = 0
x3 + x5 = 0
x4 = 0

The general solution is

x1 = − s − t x2 = s x3 = −t x4 = 0 x5 = t
The trivial solution is obtained by putting s = t = 0

Note. For a general homogeneous system, the system

corresponding to the reduced row-echelon augmented
matrix is also homogeneous
48
In particular if m < n then r < n,
where r is the number of nonzero equations in the
system corresponding to the reduced row-echelon
augmented matrix
Then we can solve for the r leading variables in
terms of the free variables and get the general
nontrivial solution
Therefore we have
Theorem. A homogeneous linear system with
more unknowns than equations has infinitely
many solutions

49
50
 1.3. MATRICES AND MATRIX
OPERATIONS

1.3.1. Matrix Notation & Terminology

1.3.2. Operations on Matrices
1.3.3. Matrix Products as Linear Combinations
1.3.4. Matrix Form of a Linear System
1.3.5. Transpose of a Matrix

51
1.3.1 MATRIX NOTATION &
TERMINOLOGY
Definition. A matrix is a rectangular array of numbers.
The numbers in the array are called the entries in the
matrix

Example. e
  − 2 
 1 2  
 



 0 1
1   1
 3 0 [2 1 0 -3]    [ 4]
  

2 
  
     3
 -1 4 0 0 0 

These are matrices of sizes 32, 14, 33, 21, 11

respectively
52
And a general m  n matrix may be written as
 a11 a12  a1n 
a 
 21 a22  a2 n 
    
 
am1 am 2  amn 
It may also be written briefly as

aij]m  n or simply aij]

The entry in row i and column j of a matrix A is also

denoted by (A)ij .

54
  b1 
b 
a = a1 a2  an] b= 2

 
bm 

Row and column matrices are denoted as vectors

An n  n matrix A is called a square matrix of order n.

The entries a11, a22, … , ann are said to be on the main

diagonal of A

55
1.3.2 OPERATIONS ON MATRICES
Definition. Two matrices are equal if they have the
same size and their corresponding entries are equal.

A=B  (A)ij = (B)ij

Example. Let
2 1  2 1   2 1 0
A=  B=  C= 
3 x  3 5  3 4 0 
Then A = B if x = 5 and A  B if x  5.
There is no value of x such that A = C since they are
of different sizes 56
Definition. Let A and B be two matrices of the same
size, then the sum A + B and the difference A – B are
matrices defined by
(A + B)ij = (A)ij + (B)ij and (A – B)ij = (A)ij – (B)ij
 2 1 0 3  -4 3 5 1
Example. Let    
   
A :=  -1 4 B :=  2 2 0 -1

0 2
  
   
 4 -2 7 0  3 2 -4 5
Then
 -2 4 5 4  6 -2 -5 2
   
   
A + B=  1 2 2 3 A − B = -3 -2 2 5
   
   
 7 0 3 5  1 -4 11 -5
57
Definition. Let A be any matrix and c is a scalar, then
the product cA is the matrix defined by
(cA)ij := c(A)ij for all i, j

cA is also called a scalar multiple of A

Example. Let  2 1 0 3
 
 
A :=  -1 0 2 4
 
 
 4 -2 7 0
Then
 4 2 0 6 -2 -1 0 -3
   
   
2A =  -2 0 4 8 (-1)A=-A =  1 0 -2 -4
   
   
 8 -4 14 0 -4 2 -7 0
58
Combining these operations we can define the linear
combination of matrices A1, A2, …, An of the same size
with coefficients (scalars) c1, c2, …, cn
n

c A = c A + c A
i =1
i i 1 1 2 2 +  + cn An
Example.

2 3 4  0 2 7 9 -6 3
A :=  
 B :=  
 C :=  

1 3 1  -1 3 -5 3 0 12

7 2 2
2 A − B + 13 C = 2 A + (−1) B + 13 C =  
 
4 3 11
59
Definition. Let A be an m  r matrix and B an r  n
matrix, then the product AB is the m  n matrix whose
entries are defined as
(AB)ij = (A)i1(B)1j + (A)i2(B)2j + … + (A)ir(B)rj

= k =r1 (A)ik(B)kj

Example. Let 4 1 4 3
 
1 2 4  
B= 0 1
A =  
 
-1 3

2 6 0  
2 7 5 2
To get (AB)23 we will multiply the corresponding entries
in row 2 and column 3 and sum up the products
60
To get (AB)23 we will multiply the corresponding entries
in row 2 and column 3 and sum up the products

4 1 4 3
1 4  

2
  
  0 -1 3 1 (AB)23=24+63+05 = 26
 
2 6 0  
2 7 5 2

The other entries are calculated in a similar manner.

The product is

12 27 30 13
AB =  

 8 -4 26 12
61
In general
 a11 a12  a1r 
a b11 b12  b1 j  b1n 
 21 a22  a2 r 
 b b  b  b 
      21 22 2j 2n 
AB =      
 ai1 ai 2  air 
 
 
    r1 r 2
b b  b  brn 

 
rj

 am1 am 2  amr 
r
Then ( AB ) ij =  aik bkj = ai1b1 j + ai 2b2 j +  + air brj
k =1

63
 1.3.4 MATRIX FORM OF
A LINEAR SYSTEM
Consider the system of m equations in n unknowns
a11 x1 + a12 x2 + + a1n xn = b1
a21 x1 + a22 x2 + + a2 n xn = b2

am1 x1 + am 2 x2 + + amn xn = bm
We can view this as an equality of the corresponding
entries of two m1 matrices

70
a11 x1 + a12 x2 +  + a1n xn   b1  The left hand side is
a x + a x +  + a x   b  the product of A with
 21 1 22 2 2n n 
=  2

      the column matrix x

    Therefore
an1 x1 + an 2 x2 +  + amn xn  bm 
a11 a12  a1n   x1   b1  This is a matrix equation
a a  a   x   b  Ax = b
 21 22 2 n  = 
2 2

    
    
an1 an 2  amn   xn  bm 
 x1   b1 
x  b 
where x=  2 and b=  2

 
   
 xn  bm  71
1.3.5 TRANSPOSE OF A MATRIX

Definition. Let A be an m  n matrix, then the

transpose of A denoted by AT is the n  m matrix that
results from interchanging the rows and columns of A
(AT)ij := (A)ji

Note. The 1st row of A becomes the 1st column of AT

The 2nd row of A becomes the 2nd column of AT, …

72
Example. The transposes of some matrices.

 a11 a12 a13 a14   a11 a21 a31 

a a32 
A = a21 a22 a23 a24  AT =  12
a22
a31 a32 a33 a34  a13 a23 a33 
 
a14 a24 a34 
 2 3
  2 5
  1
B =  
B= 1 4
T
  
  3 4 6
 5 6
 1
  T
T   D = [ 4]
C = [1 3 5] C = 3 D = [ 4]
 
 
 5
73
Definition. Let A be a square matrix of order n, then
the trace of A denoted by tr(A) is defined to be the
sum of the entries on the main diagonal of A
tr(A) = (A)11+ (A)22 + … + (A)nn

Example. The traces of some matrices.

 -1 2 7 0
 a11 a12 a13  



A = a21 a22 a23 
 3 5 -8 4
B =  

 1 2 7 -3
a31 a32 a33  



 4 -2 1 0

tr(A) = a11+ a22 + a33 tr(B) = –1 + 5 + 7 + 0 = 11

74
1.4. INVERSES; RULES OF
MATRIX ARITHMETIC

1.4.1. Properties of Matrix Operations

1.4.2. Zero and Identity Matrices
1.4.3. Inverse of a Matrix
1.4.4. Polynomial Expressions involving Matrices
1.4.5. Properties of the Transpose

75
1.4.1 PROPERTIES OF MATRIX
OPERATIONS

Theorem. Assuming the following matrix operations

are defined, then we have
(a) A + B = B + A (Commutativity of + )
(b) A + (B + C) = (A + B) + C (Associativity of + )
(c) A(BC) = (AB)C (Associativity of  )
(d) A(B + C) = AC + BC (Left Distributivity)
(e) (B + C)A = BA + CA (Right Distributivity)
76
Note. If a, b are scalars, we have similar rules
(f) a(B + C) = aB + aC
(g) (a + b) C = aC + bC
(h) a(bC) = (ab)C
(i) a(BC) = (aB)C

Recall that –B = (–1)B. Hence we also have

A(B – C) = AC – BC
(B – C)A = BA – CA
a(B – C) = aB – aC
(a – b) C = aC – bC 77
  1 2
Example. Let    4 3  1 0
 
A= 3 4 B =  
 C =  

   2 1  2 3
 
 0 1
Then  8 5
   10 9
 
AB =  20 13 and BC =  

   4 3
 
 2 1
Thus  8 5  18 15
   1 0  
     
(AB)C =  20 13   =  46 39
   2 3  
   
 2 1  4 3
and
 1 2  18 15
   
   10 9  
A(BC) =  3 4   =  46 39
      = (AB)C
   4 3
 
 0 1
 4 3 78
Note. Due to the associativity of the addition and
the multiplication we may write

A + (B + C) = (A + B) + C = A + B + C
A(BC) = (AB)C = ABC

Thus we can operate on the expressions involving

matrices almost as in the case of real numbers
except interchanging the order of the factors in a
product because the product of two matrices is not
commutative as shown in the next example

79
  -1 0  1 2
Example. Let A =  
 and B =  

 2 3  3 0

Then
 -1 -2  3 6
AB =  
  BA =  

 11 4  -3 0

❖ It may happen that AB  BA because one product

is defined but the other is not, e.g. A is a 2  3 matrix
and B is a 3  4 matrix.
❖ In the case both products are defined but having
different sizes, e.g. A is a 2  3 matrix and B is a 3
 2 matrix, then AB  BA
80
1.4.2 ZERO AND IDENTITY MATRICES

In the addition of real numbers, the number 0 plays a

neutral role. The zero matrix plays a similar role
The zero matrix of size m  n, denoted by 0m  n or
simply 0, has all of its entries equal to 0
0 
0 0 0  0 
0 0 0  0 0 0 0   
  0 0 0 0  0 
0 0 0    
0 
The zero matrix is the neutral element for addition:
A+0=0+A=A
81
The zero matrix also satisfies:
A0 = 0 0A = 0

Similarly the neutral element for multiplication is

the identity matrix I:
AI = A and IA = A
where the identity matrix In or simply I is the
square matrix of order n with 1's on the main
diagonal and 0's everywhere else

1 0 0
1 0
I 3 = 0 1 0 I2 =  
0 0 1  0 1 
82
 Example.

1 0  a11 a12 a13   a11 a12 a13 


0 1  a a  = a a 
   21 22 23 
a  21 22 a23 

1 0 0
 a11 a12 a13     a11 a12 a13 
a a =  
a23  
0 1 0 
 21 22 0 0 1 a21 a22 a23 

Theorem. Let A be a square matrix of order n. If A is

in reduced row-echelon form and has no zero row, then
A is the identity matrix In
83
 1.4.3 INVERSE OF A MATRIX

Recall that the multiplication of real numbers satisfies

the Cancellation Law, namely if a  0, then
ab = ac  b = c
or equivalently
ab = 0  b = 0
However the multiplication of matrices does not satisfy
the Cancellation Law as shown in the next example

85
 0 1  3 7 
Example. A =    0 and B =    0
 0 2  0 0 
but
0 1   3 7 
AB =     =0
0 2   0 0 
❖ We can look at the Cancellation Law for
multiplication of real numbers in another way,
namely if a is invertible (has an inverse), then
ab = 0  b = 0
The same result also holds for matrix multiplication
Let's define what is an invertible matrix first
86
Definition. A square matrix A is said to be invertible if
there is a matrix of same size B such that
AB = BA = I
B is called the inverse of A, and denoted by A–1.
A square matrix without inverse is called a singular
matrix
3 5  2 − 5
Example. B =   is an inverse of A =  
1 2  − 1 3 
 2 − 5 3 5 1 0
since AB =     =  =I
− 1 3  1 2 0 1

3 5  2 − 5 1 0
and BA =     =  =I
1 2 − 1 3  0 1 87
Now if A is an invertible matrix, then
AB = 0  B = 0

Let's assume indeed AB = 0

Then A–1(AB) = A–1 0 = 0

On the other hand

A–1(AB) = (A–1A)B = IB = B

Hence B=0
88
The above result may also be rephrased as follows:

Assume that A is a square matrix, and there is a

nonzero matrix B such that AB = 0. Then A is a
singular matrix.

1 4 0 
Example. The matrix A = 2 5 0
is singular since 3 6 0

1 4 0  0  0 
AB = 2 5 0 0 = 0 while B = 0  0
3 6 0 1 1
89
 Theorem. If A and B are invertible matrices of the
same size, then AB is invertible and

(AB)–1 = B–1A–1
Proof.
(AB)(B–1A–1) = A(BB–1)A–1 = AIA–1 = AA–1 = I
Similarly
(B–1A–1)(AB) = I
Corollary. If A1, A2, …, An are invertible matrices of
the same size, then

(A1A2…An)–1 = An–1 An–1–1… A2–1 A1–1

90
1.4.4. POLYNOMIAL EXPRESSIONS
INVOLVING MATRICES
A0 := I
If n is a positive integer, and A a square matrix, then
An := A A … A
n
If r, s are non-negative integers, then
Ar As = Ar+s
If A is also invertible and n is a positive integer, then
A–n := A–1 A–1 … A–1 = (An)–1
n
91
If A is an invertible matrix and k a nonzero scalar, then
kA is also invertible and
1 −1
−1
(kA ) = A
k
1 2 −1  3 − 2
Example. Let A =   then A =  
1 3 − 1 1 
1 2  1 2  1 2  11 30 
A =
3
     = 
1 3  1 3  1 3   15 41 

 3 −2   3 −2   3 −2   41 −30 
−3 −1 3
A = (A ) =       = 
 −1 1   −1 1   −1 1   −15 11 

11 30   41 −30 
−3
A A =
3
   = I2
15 41  −15 11  92
Let A be a square matrix of order m, and
p(x) = a0 + a1x + … + anxn
is a polynomial of degree n, then we define
p(A) = a0I + a1A + … + anAn
where I is the identity matrix of order m
 − 1 2
Example. Let A=  and p(x) = 2x2 –3x + 4
 0 3 

then p( A) = 2 A2 − 3 A + 4 I
 −1 2   −1 2  1 0  = 9 2 
2

= 2  −3  + 4  0 13
 0 3  0 3  0 1   
93
1.4.5. PROPERTIES OF THE
TRANSPOSE
Theorem. Assuming the following matrix operations
are defined, then we have
(a) (AT )T = A
(b) (A + B)T = AT + BT
(c) (A – B)T = AT – BT
(d) (kA)T = kAT
(e) (AB)T = BTAT
More generally
(A1A2…An)T = AnT An–1T… A2T A1T
94
Theorem. If A is an invertible matrix, then AT is also
invertible and
(AT )–1 = (A–1)T

− 5 − 3
Example. Let A =  
2 1
 − 5 2 1
−1 3
then A = 
T
 and A = 
− 3 1  − 2 − 5

We have 1 − 2
(A )
−1 T
=
−  =( )
AT −1

3 5 

95
1.5. FINDING INVERSES BY
ELEMENTARY MATRICES

1.5.1. Elementary Matrices

1.5.2. A Method for Inverting Matrices

96
1.5.1 ELEMENTARY MATRICES
Definition. A square matrix of order n is elementary if
it can be obtained from the identity matrix In by
performing a single elementary row operation

Example. The following matrices are elementary

1 0 0 0
1 0 3 1 0 0
1 0  0 0 0 1 0 1 0  0 1 0 
0 − 3     
  0 0 1 0
  0 0 1 0 0 1
0 1 0 0
97
 Theorem. If an elementary matrix E results from
performing a certain row operation on Im and if A is an
m  n matrix, then EA also results from performing the
same row operation on A.
 1 0 2 3 1 0 0
Example. Let  
A = 2 − 1 3 6 and E = 0 1 0
1 4 4 0 3 0 1
the elementary matrix that results from adding three
times the 1st row of I3 to the 3rd row, then
by adding three times the 1st row 1 0 2 3
of A to the 3rd row, we obtain EA = 2 − 1
 3 6
4 4 10 989
Remark. If an elementary row operation is applied to
the identity matrix I to produce an elementary matrix
E, then there is a second row operation that, when
applied to E, produces I back again
This is the inverse operation of the given operation

Row operation Inverse Row operation

I →E E →I
Multiply row i by c  0 Multiply row i by 1/c

Interchange rows i and j Interchange rows i and j

Add c times row i to row j Add –c times row i to row j

99
Example. 1 0 1 0 1 0
0 1  → 0 7  → 0 1 
     
1
multiply row 2 by 7 multiply row 2 by 7

1 0 0 1 1 0
0 1 → 1 0 → 0 1
     
interchange rows 1, 2 interchange rows 1, 2

1 0 1 5 1 0
 0 1  →  0 1 →  0 1 
     
Add 5 times row 2 to row 1 Add –5 times row 2 to row 1
100
Theorem. Every elementary matrix is invertible, and
the inverse is also an elementary matrix

Proof. Let E be an elementary matrix corresponding to

the row operation T, and T–1 the inverse row operation
of T. Let E0 be the elementary matrix obtained by
performing T–1 on the identity matrix I
It is clear that E0E is obtained by performing T on I
then performing T–1. The result is, of course I, i.e.
E0E = I
Similarly EE0 = I
Therefore E is invertible and E–1 = E0 is also an
elementary matrix
101
Theorem. Let A be a square matrix, then the
following statements are equivalent
(a) A is invertible
(b) Ax = 0 has only the trivial solution
(c) the reduced row-echelon form of A is I
(d) A is a product of elementary matrices

The matrices A and B = Em Em –1 … E1 A are said to be

row equivalent. Hence (a) and (c) show that

A matrix is invertible if and only if it is row equivalent

to the identity matrix I
104
1.5.2. A METHOD FOR INVERTING
MATRICES

Let A be invertible, then the above Theorem says that by

performing a sequence of elementary row operations T1,
T2, … , Tm we can reduce A to the identity matrix I. Let
E1, E2, … , Em be the corresponding elementary
matrices, then we have
Em Em –1 … E1 A = I
The above equality may be rewritten as
A–1 = Em Em –1 … E1 I

105
Thus we have proved

Method. To find the inverse of an invertible matrix A,

we find a sequence of elementary row operations that
reduces A to the identity matrix I, then
perform this sequence of operations to I to obtain A–1

1 2 3
Example. Find the inverse of A = 2 5 3
1 0 8

Solution. To find the inverse by the above method, we

will adjoin I to the right of A to obtain the matrix
[AI] 107
Now applying elementary row operations to [ A  I ]
to reduce the left side to the identity matrix I, then the
matrix will become
[ I  A–1 ]
The steps of the reduction are as follows
1 2 3  1 0 0  1 2 3  1 0 0
 2 5 3  0 1 0 0 1 − 3  − 2 1 0 
   
1 0 8  0 0 1 0 − 2 5  − 1 0 1

1 2 3  1 0 0 1 2 3  1 0 0
0 1 − 3  − 2 1 0  0 1 − 3  − 2 1 0 
   
0 0 − 1  − 5 2 1 0 0 1  5 − 2 − 1
108
1 2 3  1 0 0 1 2 0  − 14 6 3
0 1 − 3  − 2 1 0  0 1 0  13 − 5 − 3
   
0 0 1  5 − 2 − 1 0 0 1  5 − 2 − 1

1 0 0  − 40 16 9 
0 1 0  13 − 5 − 3
 
0 0 1  5 − 2 − 1

− 40 16 9 
A−1 =  13 − 5 − 3
 5 − 2 − 1
109
Note. If we do not know in advance that A is invertible,
then we can still carry out the foregoing reduction
steps on [ A  I ] until:
▪ The left side is I so that A is invertible and the right
side is A–1
▪ The left side contains a zero row and we conclude that
A is not invertible 1 6 4
Example. Consider the matrix A =  2 4 − 1
 
− 1 2 5 
Apply the reduction steps on [ A  I ]
1 6 4  1 0 0
0 − 8 − 9  − 2 1 0 Hence A is not
  invertible
0 0 0  − 1 1 1
110
Application. We have already seen that the following
matrix A is invertible
1 2 3
A = 2 5 3
1 0 8
Hence the corresponding homogeneous linear
system has only the trivial solution
x1 + 2 x2 + 3x3 = 0
2 x1 + 5 x2 + 3x3 = 0
x1 + 8 x3 = 0

111
 1.6. SYSTEMS OF EQUATIONS
AND INVERTIBILITY

1.6.1. Solving Linear Systems by Matrix Inversion

1.6.2. Properties of Invertible Matrices
1.6.3. Conditions for Consistency

112
1.6.1. SOLVING LINEAR SYSTEMS BY
MATRIX INVERSION
Theorem. Every system of linear equations either has
no solution, exactly one solution, or infinitely many
solutions.

Proof. Assume that the system Ax = b has at least two

solutions x1 and x2. Then A(x1–x2)=Ax1–Ax2= b–b = 0
Hence x0 = x1 – x2 is a nonzero solution of Ax = 0
Let k be any scalar, then
A(x1 + kx0) = A(x1) + kA(x0) = b + 0 = b
When k takes all real values, then the solutions x1+kx0
are all distinct and form an infinite set 113
Theorem. If A is an invertible n  n matrix, then for
each n  1 matrix b, the system of linear equations Ax
= b has exactly one solution, namely x = A–1b

Example. Let
x1 + 2 x2 + 3x3 = 5
2 x1 + 5 x2 + 3x3 = 3
x1 + 8 x3 = 17
Or putting in matrix form Ax = b, where
1 2 3  x1  5
A = 2 5 3 x =  x2  b =  3 
1 0 8  x3  17 
114
We already compute the inverse of A
− 40 16 9 
A−1 =  13 − 5 − 3
 5 − 2 − 1

Therefore
− 40 16 9   5   1 
x = A−1 Ax = A−1b =  13 − 5 − 3  3  = − 1
 5 − 2 − 1 17  2 

or x1 = 1, x2 = –1, and x3 = 2

115
❖ Consider now k systems of linear equations
Ax = b1
Ax = b2
…
Ax = bk
with the common invertible coefficients matrix A.
The solutions are
x1 = A–1b1
x2 = A–1b2
…
xk = A–1bk
These solutions may be computed using one
matrix inversion and k matrix multiplications
116
However this method is not efficient
An efficient method is to augment the coefficient
matrix A by k columns that are the right hand sides
of the systems
[ A | b1| b2| …| bk ]

Next, we reduce this matrix to the reduced row-

echelon form and separate it into k reduced systems to
obtain the solution for each system.
Another advantage of this method is that it also works
even if A is not invertible

117
Example. Solve the systems

x1 + 2 x2 + 3x3 = 4 x1 + 2 x2 + 3x3 = 1
2 x1 + 5 x2 + 3x3 = 5 and 2 x1 + 5 x2 + 3x3 = 6
x1 + 8 x3 = 9 x1 + 8 x3 = − 6

Solution. Augment the common coefficient

matrix by the right hand sides of two systems

1 2 3 4 1 
A = 2 5 3 5 6 
1 0 8 9 − 6

118
Reduce this augment matrix to reduced row-
echelon form
1 0 0 1 2 
0 1 0 0 1 
 
0 0 1 1 −1
This leads to two systems

x1 = 1 x1 = 2
x2 = 0 and x2 = 1
x3 = 1 x3 = − 1
The solutions are x1 = 1, x2 = 0, x3 = 1;
and x1 = 2, x2 = 1, x3 = –1 respectively
119
1.6.2. PROPERTIES OF INVERTIBLE
MATRICES
Theorem. Let A be a square matrix, then A is
invertible if one of the following conditions holds
(a) There exists a matrix B such that BA = I
(b) There exists a matrix B such that AB = I
Moreover the matrix B in (a) or (b) is precisely A–1
Proof. Assume that (a) holds and x is any solution of
the homogeneous linear system Ax = 0. Then
x = BAx = B0 = 0
Therefore A is invertible and B = BAA–1 = IA–1 = A–1
Similarly if (b) holds then A = B–1. Hence B = A–1
120
Corollary. If A is an n  n matrix, then A is
invertible if and only if the system Ax = b is
consistent for every n  1 matrix b.

Proof. Let's assume that the system Ax = b is

consistent for every n  1 matrix b.
Applying this for b = e1, e2, …, en, the columns of In
we obtain n column matrices x1, x2, …, xn such that
Ax1 = e1, Ax2 = e2, …, Axn = en
Let B be the matrix with columns x1, x2, …, xn, then
AB = [Ax1| Ax2| … | Axn] = [e1| e2| … | en] = I
Hence A is invertible
121
 1.6.3. CONDITIONS FOR
CONSISTENCY

Problem. Let A be a fixed m  n matrix, find all n  1

matrix b such that the system Ax = b is consistent.

If A is invertible then all n  1 matrix b satisfy the

problem.

However if A is not invertible, or A is not a square

matrix, then the n  1 matrix b must satisfy certain
conditions.

122
Example. Find conditions on b1, b1, …, bn such that
x1 + x2 + 2 x3 = b1
x1 + x3 = b2
2 x1 + x2 + 3x3 = b3
is consistent
Solution. The augmented matrix is
1 1 2 b1 
A = 1 0 1 b2 
2 1 3 b3 

123
The reduced row-echelon form is
1 1 2 b1 
0 1 1 b − b 
 1 2 
0 0 0 b3 − b2 − b1 
The 3rd row may be written as
0x1 + 0x2 + 0x3 = b3 – b2 – b1

i.e. b3 = b1 + b2
In other words, the column matrix b has the form
 b1 
 b 
 2 
b1 + b2 
124
Example. Find conditions on b1, b1, …, bn such that
x1 + 2 x2 + 3x3 = b1
2 x1 + 5 x2 + 3x3 = b2
x1 + 8 x3 = b3
is consistent
Solution. The augmented matrix is
1 2 3 b1 
2 5 3 b 
 2

1 0 8 b3 

125
The reduced row-echelon form is
1 0 0 − 40b1 + 16b2 + 9b3 
0 1 0 13b − 5b − 3b 
 1 2 3 

0 0 1 5b1 − 2b2 − b3 

There is no restriction on b1, b2, b3. The given
system has a unique solution for all b

x1 = –40b1 + 16b2 + 9b3

x2 = 13b1 – 5b2 – 3b3
x1 = 5b1 – 2b2 – b3

126
1.7. DIAGONAL, TRIANGULAR &
SYMMETRIC MATRICES

1.7.1. Diagonal Matrices

1.7.2. Triangular Matrices
1.7.3. Symmetric Matrices

127
1.7.1 DIAGONAL MATRICES
Definition. A square matrix is said to be a diagonal
matrix if all entries off the main diagonal are zero

Example. 6 0 0 0
0 − 4 0 0
1 0 0 
2 0  0 1 0  0 0 0 0
 0 − 5    
  0 0 1 0 0 0 8

128
A general n  n diagonal matrix may be written as
d1 0  0 
0 d  0 
D= 2 
  
 
 0 0  dn 
This matrix is invertible if and only all diagonal
entries are non zero. Moreover
1 0  0 
 d1 
 0 1  0 
D −1 =  d2 
    
 1 
 0 0 
d n  129
Powers of a diagonal matrix are simple to compute
d1k 0  0 
 k 
 0 d2  0 
D =
k
    
 k
 0 0  d n 
The product of a matrix with a diagonal matrix is
also simple to compute

AD = [d1c1| d2c2| … | dncn]

where c1 , c2, …, cn are column vectors of A

132
  d1r1 
d r 
And DA =  2 2  where r1, r2, …, rn are row vectors of A
  
 
d nrn 
1 5 9 1 10 27 
2  1 0 0   
 6 1    2 12 3 
Example.  0 2 0 =
3 7 2 3 14 6 
  0 0 3  
4 8 3 4 16 9 

 1 0 0  1 2 3 4   1 2 3 4 
0 2 0  5 6 7 8  = 10 12 14 16 
    
0 0 3 9 1 2 3   27 3 6 9 
133
1.7.2 TRIANGULAR MATRICES
Definition. A square matrix is a lower (upper)
triangular matrix if all entries above (below) the
main diagonal are zero. A matrix is a triangular if it is
either lower triangular or upper triangular
Example.
 a11 0  0  a11 a12  a1r 
a 0 
 21 a22  0   a22  a2 r 
         
   
 an1 an 2  anr  0 0  air 

134
Note. Let A = [aij] be a square matrix, then
▪ A is lower triangular if aij = 0 for i<j
▪ A is upper triangular if aij = 0 for i>j
Theorem.
(a) The transpose of a lower (upper) triangular matrix
is an upper (lower) triangular matrix
(b) The product of lower (upper) triangular matrices is
a lower (upper) triangular matrix
(c) The inverse of a lower (upper) triangular matrix is a
lower (upper) triangular matrix
(d) A triangular matrix is invertible if and only if its
diagonal entries are all nonzero
135
Example. Let

1 3 − 1 3 − 2 2 
A = 0 2 4  and B = 0 0 − 1
0 0 5  0 0 1 

Then

1 − 32 75  3 − 2 − 2
−1  1 2
A = 0 2 − 5  and AB = 0 0 2 
0 0 1 
5 
0 0 5 

136
1.7.3 SYMMETRIC MATRICES

Definition. A square matrix is symmetric if AT = A

Example. The following matrices are symmetric

d1 0 0 0 
1 4 5  0 d 0 0 
7 4  4 − 3 0  2 
4 − 3    0 0 d3 0 
  5 0 7  
 0 0 0 d4 
Note. A matrix is symmetric if any pair of entries
symmetric with respect to the main diagonal are
equal to each other 137
Note. Let A = [aij] be a square matrix, then A is
symmetric if and only if aij = aji for all pair (i , j)
Theorem. If A and B are symmetric matrices with the
same size, and k is any scalar, then
(a) AT is symmetric
(b) A + B and A – B are symmetric
(c) kA is symmetric
Note. The product of two symmetric matrices is not
necessary symmetric
Let indeed A and B be symmetric matrices, then
(AB)T = BTAT = BA
Therefore AB is symmetric if and only if AB = BA
(we say that A and B commute )
138
 Example. Consider the product of two symmetric
matrices that is not symmetric
1 2   − 4 1   − 2 1 
 2 3  1 0  =  − 5 2
    
The reason is that these two matrices do not commute
− 4 1 1 2 − 2 − 5 1 2 − 4 1
 1 0  2 3 =  1     1 0
    2   2 3  
On the other hand the product of the following two
symmetric matrices is symmetric since they commute
1 2 − 4 3  2 1
2 3  3 − 1 = 1 3
     139
Theorem. If A is an invertible symmetric matrix, then
A–1 is also symmetric

Proof. Since A = AT we have

(A–1)T = (AT )–1 = A–1
Note. Many matrices that arise in applications have
the form AAT or ATA where A is an m  n matrix.
It is clear that these products are square matrices of
size m  m and n  n. Moreover, they are symmetric
(AAT)T = (AT )TAT = AAT
and (ATA)T = AT (AT )T = ATA
Note that both products are invertible if A is invertible
140
 Example. The transpose of the matrix
1 3
1 − 2 4  − 2 0 
3 0 − 5 is  
   4 − 5
We have
1 3
1 − 2 4     21 − 17
3 0 − 5 − 2 0  = − 17 34 
   4 − 5  
 
and
1 3  10 − 2 − 11
− 2 0  1 − 2 4  =  − 2 4 − 8 
  3 0 − 5  
 4 − 5   − 11 − 8 41 
 
Note that both products are symmetric
141