11.

FINDING A−1 USING ELEMENTARY MATRICES 85

11. Finding A−1 Using Elementary Matrices

Before we establish the main results of this section, we recall the reader
of the following method of mathematical proofs. To say that statements
p1 , p2 , · · · , pn are all equivalent means that either they are all true or all
false. To prove that they are equivalent, one assumes p1 to be true and
proves that p2 is true, then assumes p2 to be true and proves that p3 is true,
continuing in this fashion, assume that pn−1 is true and prove that pn is true
and finally, assume that pn is true and prove that p1 is true. This is known
as the proof by circular argument.
Now, back to our discussion of inverses. The following result establishes
relationships between square matrices and systems of linear equations. These
relationships are very important and will be used many times in later sections.

Theorem 11.1
If A is an n × n matrix then the following statements are equivalent.
(a) A is invertible.
(b) Ax = 0 has only the trivial solution.
(c) A is row equivalent to In .
(d) rank(A) = n.

Proof.
(a) ⇒ (b) : Suppose that A is invertible and x0 is a solution to Ax = 0. Then
Ax0 = 0. Multiply both sides of this equation by A−1 to obtain A−1 Ax0 =
A−1 0, that is x0 = 0. Hence, the trivial solution is the only solution.
(b) ⇒ (c) : Suppose that Ax = 0 has only the trivial solution. Then the
reduced row-echelon form of the augmented matrix has no rows of zeros or
free variables. Hence it must look like
⎡ ⎤
1 0 0 ... 0 0
⎢ 0 1 0 ... 0 0 ⎥
⎢ ⎥
⎢ 0 0 1 ... 0 0 ⎥
⎢ . . . .. .. ⎥
⎢ . . . ⎥
⎣ . . . . . ⎦
.
0 0 0 .. 1 0

If we disregard the last column of the previous matrix we can conclude that
A can be reduced to In by a sequence of elementary row operations, i.e. A
is row equivalent to In .
86 MATRICES

(c) ⇒ (d) : Suppose that A is row equivalent to In . Then rank(A) =

rank(In ) = n.

(d) ⇒ (a) : Suppose that rank(A) = n. Then A is row equivalent to In . That

is In is obtained by a finite sequence of elementary row operations performed
on A. Then by Theorem 10.1, each of these operations can be accomplished
by premultiplying on the left by an appropriate elementary matrix. Hence,
obtaining
Ek Ek−1 . . . E2 E1 A = In ,
where k is the necessary number of elementary row operations needed to re-
duce A to In . Now, by Theorem 10.2, each Ei is invertible. Hence, Ek Ek−1 . . . E2 E1
is invertible and A−1 = Ek Ek−1 . . . E2 E1

Using the definition, to show that an n × n matrix A is invertible we find

a matrix B of the same size such that AB = In and BA = In . The next
theorem shows that one of these equality is enough to assure invertibilty.

Theorem 11.2
If A and B are two square matrices of size n × n such that AB = In then
BA = In and B −1 = A.

Proof
Suppose that Bx = 0. Multiply both sides by A to obtain ABx = 0. That
is, x = 0. This shows that the homogenenous system Bx = 0 has only the
trivial solution so by Theorem 11.1 we see that B is invertible, say with in-
verse C. Hence, C = In C = (AB)C = A(BC) = AIn = A so that B −1 = A.
Thus, BA = BB −1 = In

As an application of Theorem 11.1, we describe an algorithm for finding

A−1 . We perform elementary row operations on A until we get In ; say that
the product of the elementary matrices is Ek Ek−1 . . . E2 E1 . Then we have

(Ek Ek−1 . . . E2 E1 )[A|In ] = [(Ek Ek−1 . . . E2 E1 )A|(Ek Ek−1 . . . E2 E1 )In ]

= [In |A−1 ]

We ask the reader to carry the above algorithm in solving the following
problems.
11. FINDING A−1 USING ELEMENTARY MATRICES 87

Example 11.1
Find the inverse of
⎡ ⎤
1 2 3
A=⎣ 2 5 3 ⎦
1 0 8
Solution.
We first construct the matrix
⎡ ⎤
1 2 3 | 1 0 0
⎣ 2 5 3 | 0 1 0 ⎦
1 0 8 | 0 0 1
Applying the above algorithm to obtain

Step 1: r2 ← r2 − 2r1 and r3 ← r3 − r1

⎡ ⎤
1 2 3 | 1 0 0
⎣ 0 1 −3 | −2 1 0 ⎦
0 −2 5 | −1 0 1
Step 2: r3 ← r3 + 2r2
⎡ ⎤
1 2 3 | 1 0 0
⎣ 0 1 −3 | −2 1 0 ⎦
0 0 −1 | −5 2 1
Step 3: r1 ← r1 − 2r2
⎡ ⎤
1 0 9 | 5 −2 0
⎣ 0 1 −3 | −2 1 0 ⎦
0 0 −1 | −5 2 1
Step 4: r2 ← r2 − 3r3 and r1 ← r1 + 9r3
⎡ ⎤
1 0 0 | −40 16 9
⎣ 0 1 0 | 13 −5 −3 ⎦
0 0 −1 | − 5 2 1
Step 5: r3 ← −r3 ⎡ ⎤
1 0 0 | −40 16 9
⎣ 0 1 0 | 13 −5 −3 ⎦
0 0 1 | 5 −2 −1
88 MATRICES

It follows that ⎡ ⎤
−40 16 9
A−1 = ⎣ 13 −5 −3 ⎦
5 −2 −1

Example 11.2
Show that the following homogeneous system has only the trivial solution.

x1 + 2x2 + 3x3 = 0
2x1 + 5x2 + 3x3 = 0
x1 + 8x3 = 0.

Solution.
The coefficient matrix of the given system is invertible by the previous ex-
ample. Thus, by Theorem 11.1 the system has only the trivial solution

The following result exhibit a criterion for checking the singularity of a square
matrix.

Theorem 11.3
If A is a square matrix with a row consisting entirely of zeros then A is
singular.

Proof.
The reduced row-echelon form will have a row of zeros. So the rank of the
coefficient matrix of the homogeneous system Ax = 0 is less than n. By
Theorem 6.1, Ax = 0 has a nontrivial solution and as a result of Theorem
11.1, the matrix A must be singular

How can we tell when a square matrix A is singular? i.e., when does the
algorithm of finding A−1 fail? The answer is provided by the following theo-
rem

Theorem 11.4
An n × n matrix A is singular if and only if A is row equivalent to a matrix
B that has a row of zeros.

Proof.
Suppose first that A is singular. Then by Theorem 11.1, A is not row equiv-
alent to In . Thus, A is row equivalent to a matrix B = In which is in reduced
11. FINDING A−1 USING ELEMENTARY MATRICES 89

echelon form. By Theorem 11.1, B must have a row of zeros.

Conversely, suppose that A is row equivalent to matrix B with a row con-
sisting entirely of zeros. Then B is singular by Theorem 11.1. Now, B =
Ek Ek−1 . . . E2 E1 A. If A is nonsingular then B is nonsingular, a contradiction.
Thus, A must be singular

The following theorem establishes a result of the solvability of linear sys-

tems using the concept of invertibility of matrices.
Theorem 11.5
An n × n square matrix A is invertible if and only if the linear system Ax = b
is consistent for every n × 1 matrix b.
Proof.
Suppose first that A is invertible. Then for any n × 1 matrix b the linear
system Ax = b has a unique solution, namely x = A−1 b.
Conversely, suppose that the system Ax = b is solvable for any n × 1 matrix
b. In particular, Axi = ei , 1 ≤ i ≤ n, has a solution, where ei is the ith
column of In . Construct the matrix

C = x1 x2 · · · xn
Then
 
AC = Ax1 Ax2 · · · Axn = e1 e2 · · · en = In .
Hence, by Theorem 11.2, A is non-singular
Example 11.3
Solve the following system by using the previous theorem
⎧
⎨ x1 + 2x2 + 3x3 = 5
2x1 + 5x2 + 3x3 = 3
⎩
x1 + 8x3 = 17
Solution.
Using Example 11.1 and Theorem 11.5 we have
⎡ ⎤ ⎡ ⎤⎡ ⎤
x1 −40 16 9 5
⎣ x2 ⎦ = ⎣ 13 −5 −3 ⎦ ⎣ 3 ⎦
x3 5 −2 −1 17
⎡ ⎤
1
= ⎣ −1 ⎦
2
90 MATRICES

Example 11.4
If P is an n×n matrix suxh that P T P = In then the matrix H = In −2P P T is
called the Householder matrix. Show that H is symmetric and H T H = In .

Solution.
Taking the transpose of H we have H T = InT − 2(P T )T P T = H. That is, H is
symmetric. On the other hand, H T H = H 2 = (In − 2P P T )2 = In − 4P P T +
4(P P T )2 = In − 4P P T + 4P (P T P )P T = In − 4P P T + 4P P T = In

Example 11.5
Let A and B be two square matrices. Show that AB is nonsingular if and
only if both A and B are nonsingular.

Solution.
Suppose that AB is nonsingular. Suppose that A is singular. Then C =
Ek Ek−1 · · · A with C having a row consisting entirely of zeros. But then
CB = Ek Ek−1 · · · (AB) and CB has a row consisting entirely of zeros (Ex-
ample 8.3). This implies that AB is singular, a contradiction.
The converse is just Theorem 9.2 (a)
11. FINDING A−1 USING ELEMENTARY MATRICES 91

Practice Problems
Problem 11.1
Determine if the following matrix is invertible.
⎡ ⎤
1 6 4
⎣ 2 4 −1 ⎦
−1 2 5

Problem 11.2
For what values of a does the following homogeneous system have a nontrivial
solution?
(a − 1)x1 + 2x2 = 0
2x1 + (a − 1)x2 = 0

Problem 11.3
Find the inverse of the matrix
⎡ ⎤
1 1 1
⎣ 0 2 3 ⎦
5 5 1

Problem 11.4
Prove that if A is symmetric and nonsingular than A−1 is symmetric.

Problem 11.5
If ⎡ ⎤
4 0 0
D = ⎣ 0 −2 0 ⎦
0 0 3
find D−1 .

Problem 11.6
Prove that a square matrix A is nonsingular if and only if A is a product of
elementary matrices.

Problem 11.7
Prove that two m × n matrices A and B are row equivalent if and only if
there exists a nonsingular matrix P such that B = P A.
92 MATRICES

Problem 11.8
Let A and B be two n×n matrices. Suppose A is row equivalent to B. Prove
that A is nonsingular if and only if B is nonsingular.

Problem 11.9
Show that a 2×2 lower triangular matrix is invertible if and only if a11 a22 = 0
and in this case the inverse is also lower triangular.

Problem 11.10
Let A be an n × n matrix and suppose that the system Ax = 0 has only
the trivial solution. Show that Ak x = 0 has only the trivial solution for any
positive integer k.

Problem 11.11
Show that if A and B are two n × n invertible matrices then A is row equiv-
alent to B.
Determinants

With each square matrix we can associate a real number called the determi-
nant of the matrix. Determinants have important applications to the theory
of systems of linear equations. More specifically, determinants give us a
method (called Cramer’s method) for solving linear systems. Also, determi-
nant tells us whether or not a matrix is invertible.
Throughout this chapter we use only square matrices.

93
94 DETERMINANTS

12. Determinants by Cofactor Expansion

The determinant of a 2 × 2 matrix
 
a11 a12
A=
a21 a22

is the number  
 a a 
|A| =  11 12  = a11 a22 − a21 a12 .

a21 a22
The determinant of a 3 × 3 matrix can be found using the determinants of
2× matrices using a cofactor expansion which we discuss next.
If A is a square matrix of order n then the minor of entry aij , denoted
by Mij , is the determinant of the submatrix obtained from A by deleting
the ith row and the j th column. The cofactor of entry aij is the number
Cij = (−1)i+j Mij .

Example 12.1
Let
⎡ ⎤
3 1 −4
A=⎣ 2 5 6 ⎦
1 4 8

Find the minor and the cofactor of the entry a32 = 4.

Solution.
The minor of the entry a32 is
 
 3 −4 
M32 
=  = 26
2 6 

and the cofactor is C32 = (−1)3+2 M32 = −26

Example 12.2
Find the cofactors C11 , C12 , and C13 of the matrix
⎡ ⎤
a11 a12 a13
⎣ a21 a22 a23 ⎦
a31 a32 a33
12. DETERMINANTS BY COFACTOR EXPANSION 95

Solution.
We have  
 a22 a23 
C11 = (−1)  
1+1  = a22 a33 − a32 a23
a32 a33 
 

1+2  a21 a23 

C12 = (−1)  = −(a21 a33 − a31 a23 )
a31 a33 
 
 a a 
C13 = (−1)1+3  21 22  = a21 a32 − a31 a22
a31 a32 
The determinant of a matrix A of order n can obtained by multiplying the
entries of a row (or a column) by the corresponding cofactors and adding the
resulting products. Any row or column chosen will result in the same answer.
More precisely, we have the expansion along row i is

|A| = ai1 Ci1 + ai2 Ci2 + · · · + ain Cin .

The expansion along column j is given by

|A| = a1j C1j + a2j C2j + · · · + anj Cnj .

Any row or column chosen will result in the same answer.

Example 12.3
Find the determinant of the matrix
⎡ ⎤
a11 a12 a13
⎣ a21 a22 a23 ⎦
a31 a32 a33

Solution.
Using the previous example, we can find the determinant using the cofactor
along the first row to obtain

|A| =a11 C11 + a12 C12 + a13 C13

     
 a22 a23   a a   

=a11   − a12  21 23  + a13  a21 a22 
a32 a33  a31 a33   a31 a32 
=a11 (a22 a33 − a32 a23 ) − a12 (a21 a33 − a31 a23 ) + a13 a21 a32 − a31 a22
96 DETERMINANTS

Remark 12.1
In general, the best strategy for evaluating a determinant by cofactor ex-
pansion is to expand along a row or a column having the largest number of
zeroes.
Example 12.4
Find the determinant of each of the following matrices.
(a) ⎡ ⎤
0 0 a13
A = ⎣ 0 a22 a23 ⎦
a31 a32 a33
(b) ⎡ ⎤
0 0 0 a14
⎢ 0 0 a23 a24 ⎥
A=⎢
⎣ 0 a32 a33
⎥
a34 ⎦
a41 a42 a43 a44
(c) ⎡ ⎤
a11 0 0 0
⎢ a21 a22 0 0 ⎥
A=⎢
⎣ a31 a32 a33 0 ⎦
⎥

a41 a42 a43 a44

Solution.
(a) Expanding along the first column we find
|A| = a31 C31 = −a31 a22 a13 .
(b) Again, by expanding along the first column we obtain
|A| = a41 C41 = a41 a32 a23 a34
(c) Expanding along the last column we find
|A| = a44 C44 = a11 a22 a33 a44
Example 12.5
Evaluate the determinant of the following matrix.
 
 2 7 −3 8 3 
 
 0 −3 7 5 1 
 
 0 0 6 7 6 

 0 0 0 9 8 

 0 0 0 0 4 
12. DETERMINANTS BY COFACTOR EXPANSION 97

Solution.
The given matrix is upper triangular so that the determinant is the product
of entries on the main diagonal, i.e. equals to −1296

Example 12.6
Use cofactor expansion along the first column to find |A| where
⎡ ⎤
3 5 −2 6
⎢ 1 2 −1 1 ⎥
A=⎢ ⎣ 2 4
⎥
1 5 ⎦
3 7 5 3

Solution.
Expanding along the first column we find

|A| = 3C11 + C21 + 2C31 + 3C41

= 3M11 − M21 + 2M31 − 3M41
= 3(−54) + 78 + 2(60) − 3(18) = −18
98 DETERMINANTS

Practice Problems
Problem 12.1
Evaluate the determinant of each of the following matrices
(a)  
3 5
A=
−2 4
(b) ⎡ ⎤
−2 7 6
A = ⎣ 5 1 −2 ⎦
3 8 4

Problem 12.2
Find all values of t for which the determinant of the following matrix is zero.
⎡ ⎤
t−4 0 0
A=⎣ 0 t 0 ⎦
0 3 t−1

Problem 12.3
Solve for x  
   1 0 −3 
 x −1   
  =  2 x −6 
 3 1−x   
 1 3 x−5 

Problem 12.4
Evaluate the determinant of the following matrix
⎡ ⎤
1 2 3
A= 4 5⎣ 6 ⎦
0 0 0

Problem 12.5
Let ⎡ ⎤
4 −1 1 6
⎢ 0 0 −3 3 ⎥
A=⎢
⎣ 4 1
⎥
0 14 ⎦
4 1 3 2
Find M23 and C23 .
12. DETERMINANTS BY COFACTOR EXPANSION 99

Problem 12.6
Find all values of λ for which |A| = 0, where
 
λ−1 0
A= .
2 λ+1

Problem 12.7
Evaluate the determinant of the matrix
⎡ ⎤
3 0 0
A = ⎣ 2 −1 5 ⎦
1 9 −4

(a) along the first column.

(b) along the third row.

Problem 12.8
Evaluate the determinant of the matrix by a cofactor expansion along a row
or column of your choice.
⎡ ⎤
3 3 0 5
⎢ 2 2 0 −2 ⎥
A=⎢⎣ 4 1
⎥
−3 0 ⎦
2 10 3 2

Problem 12.9
Evaluate the determinant of the following matrix by inspection.
⎡ ⎤
1 2 7 −3
⎢ 0 1 −4 1 ⎥
A=⎢ ⎣ 0 0 2
⎥
7 ⎦
0 0 0 3

Problem 12.10
Evaluate the determinant of the following matrix.
⎡ ⎤
sin θ cos θ 0
A = ⎣ − cos θ sin θ 0 ⎦
sin θ − cos θ sin θ + cos θ 1
100 DETERMINANTS

Problem 12.11
Find all values of λ such that |A| = 0.
(a)
 
λ − 1 −2
A= ,
1 λ−4

(b)
⎡ ⎤
λ−6 0 0
A=⎣ 0 λ −1 ⎦
0 4 λ−4
13. EVALUATING DETERMINANTS BY ROW REDUCTION 101

13. Evaluating Determinants by Row Reduc-

tion
In this section we provide a simple procedure for finding the determinant of a
matrix. The idea is to reduce the matrix into row-echelon form which in this
case is a triangular matrix. Recall that a matrix is said to be triangular if
it is upper triangular, lower triangular or diagonal. The following theorem
provides a formula for finding the determinant of a triangular matrix.
Theorem 13.1
If A is an n × n triangular matrix then |A| = a11 a22 . . . ann .
Example 13.1
Compute |A|.
(a)
⎡ ⎤
1 2 3
A = ⎣ 0 4 5 ⎦,
0 0 6
(b)
⎡ ⎤
1 0 0
A = ⎣ 2 3 0 ⎦,
4 5 6
Solution.
(a) Since A is triangular, |A| = (1)(4)(6) = 24.
(b) |A| = (1)(3)(6) = 18
Example 13.2
Compute the determinant of the identity matrix In .
Solution.
Since the identity matrix is triangular with entries equal to 1 on the main
diagonal, |In | = 1

The following theorem is of practical use. It provides a technique for eval-

uating determinants by greatly reducing the labor involved. We shall show
that the determinant can be evaluated by reducing the matrix to row-echelon
form.