Chapter 6

The determinant of a square matrix

6.1 Definition of the determinant and properties

In this chapter, and in order to make the life of our students easier, we will content ourselves to a
simple and practical notion of a determinant. But this definition will complicate the proofs of two of
the properties of determinant. They will be left as exercises in the exercise session.

6.1.1 Definition

 
The determinant of a square matrix A  a ij 1i  n , denoted by det(A) or A , is the real number
1 j n
defined by induction on the dimension (or the order) n of the matrix A as follows:

 First case:

If n  1, the matrix A is of the form: A   a11  . The determinant of A is by definition a11 , i.e.
det( a11 )  a11 .

Examples

3  3 ; π  π ; 0  0 ;  2   2 ;......

 Second case

a a 
If n  2 , the matrix A is of the form  11 12  . We define the cofactor of the element a ij of A
 a 21 a 22 
as being the product of the sign  1i  j by the minor of the element a ij , i.e. by the determinant of

the 11 submatrix obtained by deleting the i th row and the jth column of the matrix A. The minor
of the element a ij is denoted by A ij . To calculate the determinant of the matrix A of dimension
2  2 , we start by choosing or by fixing a row or a column according to which we will ‘‘develop’’
the determinant. Thus,

 a11 a12  " developed " according


a 21 a 22 to the first
[Link] cofactor  a12 . its cofactor
row that we have chosen

11 1 2 a a a11 a12

 a11.  1 . A11  a12 .  1 . A12  a11.(1)11. 11 12  a12 . (1)1 2 .
a 21 a 22 a 21 a 22
its cofactor its cofactor

 a11.a 22  a 21.a12 .

92
Example 1

2 4
 2.(5)  3.4  22 .
3 5

Example 2

2 3
 2.(9)  6.(3)  0.
6 9

 Third case

 a11 a12 a13 

 
If n  3 , the matrix A is of the form  a 21 a 22 a 23  . We define the cofactor of the element a ij
a a 33 
 31 a 32
of A as being the product of the sign  1i  j by the minor of the element a ij , i.e. by the

determinant of the 2  2 submatrix obtained by deleting the i th row and the jth column of the matrix
A. The minor of the element a ij is denoted by A ij . To calculate the determinant of the matrix A of
dimension 3  3 , we start by choosing or by fixing a row or a column according to which we will
‘‘develop’’ the determinant. Thus,

a11 a12 a13 " developed " according to

a 21 a 22 a 23
rd

 a 31 a 32 a 33  the 3 row that we have chosen

a31 . its cofactor  a32. its cofactor  a33. its cofactor

31 3 2 3 3
 a 31.  1 . A31  a 32 .  1 . A32  a 33.  1 . A33 
its cofactor its cofactor its cofactor

a11 a12 a13 a11 a12 a13 a11 a12 a13

31 3 2 3 3
 a 31.  1 . a 21 a 22 a 23  a 32 .  1 . a 21 a 22 a 23  a 33.  1 . a 21 a 22 a 23
a 31 a 32 a 33 a 31 a 32 a 33 a 31 a 32 a 33

31 a a13 3 2 a11 a13 3 3 a a

 a 31.  1 . 12  a 32 .  1 .  a 33.  1 . 11 12
a 22 a 23 a 21 a 23 a 21 a 22

 a31(a12.a 23  a 22.a13 )  a32 (a11.a 23  a 21.a13 )  a33 (a11.a 22  a21.a12 ) .

93
Example

2 1 4 "developed " according

 3 0 2   nd
3. its cofactor  0. its cofactor 2. its cofactor
4 1 5 to the 2 row

2 1 4 2 1 4 2 1 4
2 1 2 2 23
 3.  1 . 3 0 2  0.  1 . 3 0 2   2  .  1 . 3 0 2
4 1 5 4 1 5 4 1 5

2 1 1 4 23 2 1
 3.  1 .  0  2.  1 .
1 5 4 1

 3.(5  4)  0  2.(2  4)  3  4  7.

 Fourth case

 a11 a12 a13 a 14 

 
a 21 a 22 a 23 a 24 
If n  4 , the matrix A is of the form  . We define the cofactor of the
 a 31 a 32 a 33 a 34 
 
 a 41 a 42 a 43 a 44 
i j
element a ij of A as being the product of the sign  1 by the minor of the element a ij , i.e. by

the determinant of the 3  3 submatrix obtained by deleting i th row and the jth column of the matrix
A. But, we already know how to calculate this 3  3 determinant.
The minor of the element a ij is denoted by A ij . To calculate the determinant of the matrix A of
dimension 4  4 , we start by choosing or by fixing a row or a column according to which we will
‘‘develop’’ the determinant. Thus, and for example,

a11 a12 a13 a 14

a 21 a 22 a 23 a 24 " developed " according
a 31 a 32 a 33 a 34 
to the 2nd column
a 41 a 42 a 43 a 44

a12 . its cofactor  a 22 . its cofactor  a32 . its cofactor  a 42 . its cofactor
1 2 2 2 3 2 4 2
 a12 .  1 . A12  a 22 .  1 . A 22  a 32 .  1 . A32  a 42 .  1 . A 42 
its cofactor its cofactor its cofactor its cofactor

94
 a11 a12 a13 a 14 a11 a12 a13 a 14
1 2 a a 22 a 23 a 24 2  2 a 21 a 22 a 23 a 24
 a12 .  1 . 21  a 22 .  1 .
a 31 a 32 a 33 a 34 a 31 a 32 a 33 a 34
a 41 a 42 a 43 a 44 a 41 a 42 a 43 a 44

a11 a12 a13 a 14 a11 a12 a13 a 14

3 2 a 21 a 22 a 23 a 24 4  2 a 21 a 22 a 23 a 24
a 32 .  1 .  a 42 .  1 .
a 31 a 32 a 33 a 34 a 31 a 32 a 33 a 34
a 41 a 42 a 43 a 44 a 41 a 42 a 43 a 44

a 21 a 23 a 24 a11 a13 a14 a11 a13 a14

1 2 2 2 3 2
 a12 .  1 . a 31 a 33 a 34  a 22 .  1 . a 31 a 33 a 34  a 32 .  1 . a 21 a 23 a 24
a 41 a 43 a 44 a 41 a 43 a 44 a 41 a 43 a 44
a11 a13 a14
4 2
 a 42 .  1 . a 21 a 23 a 24 .
a 31 a 33 a 34

Therefore the calculation of a 4  4 determinant is reduced to the calculation of 4 determinants of

dimension 3  3 , and we already know how to calculate the determinants of order 3  3 .

 Fifth case

If n  5 , we continue with the same reasoning: the calculation of a determinant of dimension n  n

is reduced to the calculation of n determinants of smaller dimension  n  1   n  1 .

6.1.2 Properties

1) The value of the determinant does not depend on the choice of the row or the column in which
it is developed. But is wise to develop according to the row or the column that contains the
most of zeros, i.e. the maximum of zero elements. Proof to be done in the exercise session.

Example

1 λ π e
Calculate the following determinant : 0 2λ 3 .
0 0 3 λ

Answer

To calculate it, it is wise to develop it according to the first column or according to the third row,
because they contain two zero elements. Thus,

95
1 λ π e " developed " according
0 2λ 3 
st
0 0 3  λ to the 1 column
11 2  λ 3
1  λ  .its cofactor  0  0  1  λ  .  1 .  1  λ  .  2  λ  . 3  λ  .
0 3 λ

Remark

We shall see later that the determinant of a triangular matrix is equal to the product of its diagonal
entries.

2) det  A.B   det  A  .det  B  . Proof to be done in the exercise session.

3) det  A  B   det  A   det  B  .

Example-proof

 1 0   1 0   1 0  1 0 
det      det    det   . In fact:
 0 1   0 1  0 1  0 1

 1 0   1 0   0 0 1 0  1 0 
det      det    0.0  0.0  0 . But det    det  
 0 1   0 1  0 0 0 1  0 1

 1.1  0.0    1 .(1)  0.0  2  0.

4)  
det t A  det  A  . Proof to be done in the exercise session.
5) The determinant of an upper or a lower triangular matrix is equal to the product of the entries
which are on the diagonal, i.e.:

 a11 a12 a1n 

 
0 a 22
det    a .a . ..... .a .
  11 22 nn
 
 0 0 a nn 

Proof

By induction on n, and by developing for example according to the first column.

6) The determinant of a diagonal matrix is equal to the product of the entries which are on the
diagonal, in other words:

96
  a11 0 0 
 
0 a 22
det    a .a . ..... .a .
 0  11 22 nn
 
 0 0 a nn 

Proof

A diagonal matrix is necessarily triangular.

7) The determinant of an identity matrix is equal to 1, i.e. :

1 0 0
 
0 1
det(Id)  det    1.1. .... .1  1 .
 0
 
0 0 1

Proof

An identity matrix is necessarily triangular and diagonal.

0 0 0
 
0 0
8) The determinant of a zero square matrix is zero, i.e. det    0.
 0
 
0 0 0

9) The determinant of an elementary matrix Eij (λ) of type 1 (see paragraph 4.1.2 of chapter 4) is
1  det (Eij (λ))  1 .

Proof

An elementary matrix of type 1 is a special case of a triangular matrix having 1’s on the diagonal.

10) The determinant of an elementary matrix Di (λ) of type 2 (see paragraph 4.1.2 of chapter 4)
is λ  det  Di (λ)   λ .

Proof

An elementary matrix of type 2 is a special case of a triangular or diagonal matrix having on the
diagonal: the term 1  n  1 _times and the term  only once.

11) The determinant of an elementary matrix Pij of type 3 (see paragraph 4.1.2 of chapter 4)
is -1  det (Pij )  1 .

97
Proof

0 1
By induction on n, and if n = 2  det (P12 )  det    1 .
1 0

The following three properties reflect the effects on the determinant of the three elementary
operations on the rows or on the columns of a matrix.

12) det (Eij (λ).A)  det(Eij (λ)).det(A)  [Link](A)  det(A) and

det ([Link] (λ))  det(A).det(Eij (λ))  det(A).1  det(A) .

Therefore, and using the properties 12 and 13 of paragraph 5.4.2, we can conclude that if we
add to a given row (respectively to a given column) a multiple of another row (respectively of
another column), or more generally, if we add to a given row (respectively to a given column) a
linear combination of the other rows (respectively the other columns), then the determinant
does not change.

Example

2 1 3 R R  2R 2 1 3 R R  3R 2 1 3
2 2 1 3 3 1
4 2 1  0 4  5  0 4 5  2.4.7  56 .
6 3 2 6 3 2 0 0 7
Triangular

13) det (Di (λ).A)  det(Di (λ)).det(A)  λ.det(A) and det ([Link] (λ))  det(A).det(Di (λ))  λ.det(A) .

Therefore, and using the properties 14 and 15 of paragraph 5.4.2, we can conclude that if we
multiply all the entries of a given row (respectively of a given column) by λ , then the
determinant will be multiplied by λ . Thus, and for example, we have:

λ.a11 λ.a12 λ.a1n a11 a12 a1n

a 21 a 22 a 2n a a 22 a 2n
 λ. 21 .

a n1 a n2 a nn a n1 a n2 a nn

Example

1 12 4 1 12 4 1 12 4 1 4 4

2 6 8  2. 2 6 8  2.2. 1 3 4  2.2.3. 1 1 4 .
4 24 2 2 12 1 2 12 1 2 4 1

Remark

There is an enormous difference between the product of a scalar by a matrix and the product of a
scalar by a determinant:

98
Example

 3 e π   6 2e 2π 
   
2.  4 1 2    8 2 4  .
12 6 1   24 12 2 
   

3 e π 6 2e 2π 3 e π 3 e 2π 6 2e 2π
But 2. 4 1 2  4 1 2  8 2 4  4 1 4  8 2 4 .
12 6 1 12 6 1 12 6 1 12 6 2 24 12 2

14) Let A be a matrix of dimension n  n , and let  be a real number. Then det(.A)  n .det(A).

Proof

 R1   .R1 
   
 R2   .R 2 
Let A  . Then the matrix .A  . By applying n times the property 13 above, we can
   
   
 Rn   .R n 
conclude that det(.A)  ..... .det(A)   n .det(A).
n _ times

15) det (Pij.A)  det(Pij ).det(A)  det(A) and det ([Link] )  det(A).det(Pij )  det(A) .
Therefore, and using the properties 16 and 17 of paragraph 5.4.2, we can conclude that if we
interchange two rows (respectively two columns), then the determinant changes its sign.

Example

1 2 1 R  R 1 2 1
2 3
0 0 4   0 3 5   (1).(3).(4)   12.
0 3 5 0 0 4
Triangular

Very useful and very important remark

Let T be the transform of a matrix A using the three elementary operations of type 1, 2 and 3 on the
rows or on the columns of A. The two matrices A and T are said to be equivalent  there are two
invertible matrices B and C such that B.A.C = T. Through the above three properties, we can
deduce that if the determinant of A is nonzero (respectively zero), then the determinant of its
transform T remains nonzero (respectively zero).

16) Using the two elementary operations of type 1 and 3 on the rows (properties 12 and 15), we
can return any determinant to the upper triangular form. Then, we easily calculate its value by
multiplying its diagonal entries.

99
Proof

If the first column is zero, then we hide the first column and the first row, and we move to the
second column. Moreover, in this case, the determinant is zero (property 18).
If the first column of the matrix is nonzero, then we can assume (by interchanging two rows if
necessary) that the first term a11 , at the top, in the first column is nonzero. Then, we perform
appropriate elementary operations on the rows to cancel all the terms of the first column that lie
a a
below a11 . More specifically, we perform R 2   R 2  21 .R1 ,….. and R n  R n  n1 .R1 .
a11 a11
 a 11 
 
 0 
Thus the first column becomes . Then hide the first row and continue this process on the
 
 
 0 
second column which is on the right. Continue this process on the columns, always from the left to
the right.

Remark

To reduce the determinant to the lower triangular form, we proceed in the same way but “from
the right to the left”.

Example

Without expanding and by using elementary operations on the rows, calculate the determinant of
 1 2 3 4
 
 2 3 4 1
the following matrix: .
 3 4 1 2
 
 4 1 2 3

Solution

Recall that the aim is to cancel the entries which lie below the diagonal and reduce the matrix to an
upper triangular form.

To make the above matrix upper triangular, the cancellation of the entries is done “from bottom to
top” or “from top to bottom” but necessarily “from the left to the right”. Here is for example a
diagram that illustrates the way forward:

1 2 3 4
2 3 4 1
 .
3 4 1 2
 
4 1 2 3

100
By performing the elementary operations on the indicated rows, we obtain:

1 2 3 4 1 2 43 1 2 3 4
2 3 4 1 R 4  R 4  4R1 2 3 1 R 3  R 3 3R1 2
4 3 4 1
3 4 1 2  3 4 1 2
 0 2 8 10
4 1 2 3 0 7 10 13 0 7 10 13

1 2 3 4 1 2 3 4
R 2  R 2  2R1 0 1 2 7 0 1 2 7
 0 2 8 10   1 .  2  .  1 0 1 4 5
0 7 10 13 0 7 10 13

1 2 3 4 1 2 3 4
R 4  R 4  7R 2 0 1 2 R 3 R 3  R 2
7 0 1 2 7
  2  .   2  .
0 0 2 2
0 1 4 5
0 0 4 36 0 0 4 36

1 2 3 4
R 4  R 4  2R 30 1 2 7
  2  . 0 0 2 2
  2  .(1).(1).(2).(40)  160.

0 0 0 40

Important remark

To avoid calculations with fractions, it is sometimes necessary to perform elementary operations of

type 2 of the form R i   .R i , or more generally of the form R i  .R i  R j . But
unfortunately, these operations modify the determinant by multiplying it by  . Then, if we perform
them, we will be obliged to divide the determinant by  , in order to avoid changing its value.

17) SARRUS Rule to compute the determinant of 3  3 matrix:

a11 a12 a13

a 21 a 22 a 23  a11.a 22 .a 33  a 21.a 32 .a13  a 31.a12 .a 23  a13 .a 22 .a 31  a 23 .a 32 .a11  a 33 .a12 .a 21 
a 31 a 32 a 33 .
a11 a12 a13
a 21 a 22 a 23

101
Example

1 2 4
2 1 3  (1).(1).(1)  (2).(0).(4)  (4).(2).(3)  (4).(1).(4)  (1).(0).(3)  (2).(2).(1) 
4 0 1
1 2 4  1  0  24   16  0  4   25  20  45.
2 1 3

18) If one of the rows (respectively of the columns) is zero, then the determinant is zero.

Proof

It suffices to expand the determinant according to this zero row (respectively according to this zero
column).

19) If the rows (respectively the columns) are not linearly independent, i.e. if one of the rows is a
linear combination of the other rows, then the determinant is zero.

Proof

We can add to this row the opposite of the linear combination of the other rows to make this row
zero, and therefore the determinant is zero. In order to set these ideas down, suppose we have:
R 3  2.R1  4.R 2  1.R 4 . Then we perform the following elementary operation:
R 3  R 3   2.R 1  4.R 2  1.R 4  , and thus the row R 3 becomes a zero row. Then, the
determinant will be zero according to the previous property.

20) If the determinant of a matrix is zero, then the rows (respectively the columns) are not linearly
independent.

Proof

Let A be a square matrix of order n. The proof is done by contradiction. If the n rows-vectors of A
were linearly independent, the dimension of the subspace generated by the n rows-vectors would be
n, and therefore the rank of the matrix would be n (see further corollary 9.3.4). Thus, the
determinant of the matrix A would be nonzero. This contradicts the hypothesis.

Example 1
 1   2   1
     
Show that the 3 vectors  2  ;  1  and  1  from a basis of 3
.
 3  0  2
     
Solution

First, let us show that these 3 vectors are linearly independent. Therefore, we calculate the
following determinant:

102
 1 2 1
2 1 1
 (1).(1).(2)  (2).(0).(1)  (3).(2).(1)  ( 1).( 1).(3)  (1).(0).(1)  (2).(2).(2) 
3 0 2
1 2 1
2 1 1
 2  6  3  8  3  0  these three vectors are linearly independent.
But their number  3  dim 3 , then they form a basis of 3
.

Example 2

1 2 3
 
Calculate the determinant of the following matrix  2 0 2  .
 3 2 1 
 
Solution

1 2 3
2 0 2  0 , because C1  C2  C3  C1  C2  C3  0 , i.e. the columns are not linearly
3 2 1
independent.

Example 3

 1 2 3 
 
Calculate the determinant of the following matrix  0 4 5  .
 2 4 6 
 
Solution

1 2 3
0 4 5  0 , because R3  2R1  2.R1  0.R 2 1.R3  0 , i.e. the rows are not linearly
2 4 6
independent.

21) For every square matrix A of order m and every matrix C of size m  n , we have:

In 0 A C
 A , and A.
C A 0 In

Proof

To prove this property, it suffices to develop the first determinant (respectively the second), n_times
according to the first row (respectively the last row).

103
22) Determinant of a block triangular matrix.
Let A be a square matrix of order m and B be a square matrix of order n and C be a matrix of
size m  n . Then:

A C
 A.B .
0 B

Proof

We can easily show that:

 A C   Im 0  A C  A C Im 0 A C
   . 0 I    .  A.B .
 0 B  0 B  n 0 B 0 B 0 In
B A

6.2 Inverse of a matrix and properties

Recall the definition of an invertible matrix (property 19 of paragraph 5.4.2): a matrix

A  Mnn ( ) is said to be invertible if there exists a matrix denoted by A 1  M nn ( ) such that
A.A1  A1.A  Id. And in fact, it is enough to have: A.A 1  Id .

6.2.1 Theorem

A matrix A  Mnn ( ) is invertible if and only if its determinant is nonzero.

Proof of the necessary condition (  )

Hypothesis: The matrix A is invertible, i.e. the inverse matrix A 1 exists with A.A1  Id.

Conclusion: det(A)  0 .

In fact, since A.A 1  Id  det(A.A1)  det(Id)  det(A).det(A1)  1  det(A)  0.

Proof of the sufficient condition (  )

Hypothesis: det(A)  0 .

Conclusion: the inverse matrix A 1 exists with A.A1  A1.A  Id.

 a11 a12 a1n 

 
a 21 a 22 a 2n 
In fact, if A   , the inverse matrix is
 
 
 a n1 a n2 a nn 

104
 t
 cofactor of a11 cofactor of a12 cofactor of a1n 
 
1 cofactor of a 21 cofactor of a 22 cofactor of a 2n 
1
A  .  .
det(A)  
 
 cofactor of a n1 cofactor of a n2 cofactor of a nn 
called the adjugate matrix of A

In fact, and using the properties of the product of two matrices and the properties of the
determinant, we will show (in the exercise session) that:
[Link] adjugate  A. t the matrix of cofactors  det(A).I 
1
A. . t the matrix of cofactors  I  A.A1  A1.A  Id.
det(A)

Example 1

1 2 1
Let the matrix A    . Is A invertible? If yes, determine its inverse A .
 3 4 

Solution

1 2
As det    4  6  10  0 , the matrix A is invertible. And thus,
3 4 

t
1 1  cofactor of 1 cofactor of 2 
A  .  
det(A)  cofactor of 3 cofactor of  4 

t  4 2 
  111 .det(4) 1 2 
 1 .det(3)  1  4 2   10
1 10 
 .    .  .
10   12 1 .det(2)  
 1
2 2
.det(1)  10  3 1   3 1 
   
 10 10 

Example 2

 1 2 3
 
Let the matrix B   2 1 0  . Is B invertible? If yes, determine its inverse B1 .
 1 3 1 
 

Solution
 1 2 3  " developed " according
 
As det  2 1 0    [Link] cofactor  [Link] cofactor  [Link] cofactor
 1 3 1  to the 2nd row
 
2 3 1 3
 2.(1)2 1.  1.(1)2  2 .  0  2.(2  9)  1.(1  3)  10  0  B is invertible 
3 1 1 1

105
 t
 cofactor of 1 cofactor of 2 cofactor of 3 
1 1  
B1  .the adjugate of B  .  cofactor of  2 cofactor of 1 cofactor of 0 
det(B) det(B)  
 cofactor of  1 cofactor of 3 cofactor of 1 

t
 11 1 0 2 0 2 1 
  1 .  11 2 .  113 . 
 3 1 1 1 1 3 
 
2 1 2 3 1 3 23 1 2
 12  2 .
1
 .   1 .  1 . 
10  3 1 1 1 1 3 
 
  131 . 2 3
 13 2 .
1 3 3 3 1 2 
 1 .
 2 0 2 1 
 1 0

t
 1 2 5   1 7 3 
1   1  
  .  7 4 5    .  2 4 6  .
10   10  
 3 6 5   5 5 5 

Verification (which is not necessary):

 1 2 3  1 7 3   10 0 0  1 0 0
1   1   1    
B.B   2 1 0  . .  2 4 6   .  0 10 0    0 1 0   Id.
 1 3 1  10  5 5 5  10  0 0 10   0 0 1 
    

6.2.2 Properties of the inverse of a matrix

   A. The proof follows from the equality:

1
1) If det(A)  0  A1 A.A1  A1.A  Id.

If det(A)  0   t A    A1  .
1 t
2)

Proof

 A.A1  
t
Apply the transpose to both sides of the following equality: A.A 1  Id  t
Id ,
Id
 
t
A 1 . t A
hence the conclusion.

1
3) det(A)  0 and det(B)  0   A.B   B1.A 1 .

Proof

 A.B .  B1.A1   A.  B.B1  .A1  [Link].A1  A.A1  Id   A.B

1
 B1.A1

106
The reciprocal is also true:

4) If the product A.B is invertible, i.e. if det(A.B) ≠ 0, the two matrices A and B are invertible.

Proof


(A.B).(A.B)-1 = A. B.  A.B 
1
  I , and (B.A).(B.A) -1

 B. A.  B.A 
1
  I.
A 1 B1

5)  I 1  I , since I.I  I.

 1 
a 0 0 
1  11 
 a11 0 0 
  If all the a ii  0  0 1 
6) The inverse 
0 a 22    a 22 
 0   .
   0 
 0 0 a nn   
 0 0
1 
 a nn 


In other words, the inverse of a diagonal matrix is also a diagonal matrix.

Example

1 
2 0 0 
1
2 0 0   
 
0 0 .
1
0 3 0   3 
 0 0 4   
  1
 0 0  
 4

 1 
a ? ? 
1  11 
 a11 a12 a1n 
  If all the a ii  0  0 1 
7) The inverse  0 a 22    a 22 
   .
   ? 
 0 0 a nn   
 0 0
1 
 a nn 


In other words, the inverse of an upper triangular matrix is also an upper triangular matrix.

107
Example

1  1 1 1 
? ?  2 
 2 3 1 
1  2 2 8 
   
 
?  .....   0
 1  1 1
 0 3 2    0 3   3
 .
6
 0 0 4     
  1 1
 0 0   0  
 0
 4  4

Remark

The proofs (will be done in the exercise session) of properties 6 and 7 derive from the inverse
formula with the cofactors method.

8) If S is a symmetric matrix, and if det(S)  0  S1 is also a symmetric matrix.

Proof

   
property 2 1
Since S is symmetric, then t S  S . Thus, we have : t S1  t S  S1. Therefore, the

matrix S1 is symmetric.

9) If A is an invertible antisymmetric matrix  A1 is an antisymmetric matrix.

Proof

Since A is antisymmetric, then t A   A . Thus, we have:

   
property 2 1 1
t
A 1  t A    A    A 1. Thus the matrix A 1 is also antisymmetric.

10) An elementary matrix E ij    of type 1 is invertible and  Eij      Eij    .

1

Proof

The inverse or the reciprocal of the operation which adds to the row Ri the row Rj multiplied by 
is the operation which adds to the row Ri the row Rj multiplied by  .

1
11) An elementary matrix D i    of type 2 is invertible, and  Di      Di   .
1


Proof

The inverse or the reciprocal of the operation which multiplies the row Ri by  is the operation
1
which multiplies the row Ri by .


108
12) An elementary matrix Pij of type 3 is invertible, and  Pij   Pij .
1

Proof

The inverse or the reciprocal of the operation which interchanges the row Ri with the row Rj is the
operation itself.

13) Let A be an invertible matrix and α be a nonzero real number. The matrix α .A is invertible
1
and ( α .A )-1 = .A 1 .


Proof

1 1 1
The result derives from the fact that: ( α .A).( .A ) = ( α . ).(A.A 1 ) = 1. I =I.
 

14) Second method to determine the inverse of an invertible square matrix A.

Under the property 16 of paragraph 6.1.2, and by using elementary operations on the rows,

we can reduce the matrix A to the upper triangular form. Let T be this upper triangular.

Since the matrix A is invertible, its determinant is nonzero. Therefore, the determinant of T is also

nonzero (since the determinants of the three types of elementary matrices are nonzeros: Remark

of the property 15 of paragraph 6.1.2. Thus the diagonal entries of the matrix T must be nonzeros,

otherwise the determinant of T and the determinant of A would be zeros.

Therefore, we can continue this process to also cancel the entries which lie above the diagonal of

the matrix T (from the rightmost column to the left). More precisely, the matrix T will be reduced to

the diagonal form.

By using, if necessary, elementary operations of type 2, we can arrive to the identity matrix.

In summary, we can reduce any invertible matrix to the identity form, using elementary operations

on the rows. Moreover, to reduce an inventible matrix to the identity form, using the elementary

operations of the rows, is simply to reduce the matrix to the reduced echelon form. This subject will

be discussed in chapter 9.

109
The following diagram explains then the second method to find the inverse:

Let A  Mnn ( ) and det(A)  0 Identity matrix Id

An elementary operation on a row The same elementary operation

An elementary operation on a row The same elementary operation

. .
. .
. .

An elementary operation on a row The same elementary operation

Identity matrix Id A 1

Proof

Let A be an invertible matrix. To carry out elementary operations on the rows of A, amounts to

multiply A on the left side by elementary matrices. Thus:

an elementary matrix. an elementary matrix. ... . an elementary matrix . A  Id .

A 1
Therefore, the inverse matrix A 1 is the product of these appropriate elementary matrices used.

Consequently, it is sufficient to apply the corresponding elementary operations on the rows of the

identity matrix to obtain the inverse matrix A 1 .

Example

Determine, using the second method above, the inverse of the matrix

1 2
A .
 3 4 

110
Solution

1 2 1 0
Let A    Id   
3 4  0 1

R 2  R 2  3R1 R 2  R 2  3R1

1 2   1 0
   
0
 10 
  3 1 

R2 R2
R2  R2 
10 10

1 2   1 0 

0 
  
 1   3  1 
 10 10 
R1  R1  2R 2
R1  R1  2R 2

 4 2 
 10 10 
1 0    A 1
   Id  3  
1
0 1 
 10 10 

Remarks

1) If the order of the matrix is greater than 2, and to cancel the entries which lie below
(respectively above) the diagonal, we proceed from the leftmost column to the right
(respectively from the right to the left).

2) For this method, we will exclusively use elementary operations on the rows or exclusively
elementary operations on the columns.

3) Despite the fact that the elementary operations of type 2 and type 3 change the value of the
determinant, we can use them, provided they are applied at the same time and simultaneously
on both sides.

4) Since A 1 is the product of elementary matrices, and since the inverse of an elementary matrix
is an elementary matrix, then every invertible matrix A is also the product of elementary
matrices. This interesting result will be used later to demonstrate in the exercise session the
properties 2 and 4 of the determinant.

111