Matrix Inverses

1 Properties of Matrix Inverses

Definition 1.1. Invertible Matrix
An n × n matrix A is invertible if there exists an n × n matrix C such that
CA = In and AC = In . C is called the inverse of A and we write C = A−1 .

Theorem 1.1. The inverse of a matrix is unique.

Proof. Let B and C be inverses of A. Then B = BI = BAC = IC = C. 

Since the inverse of a matrix is unique, it is appropriate to refer to the inverse

of A, as in there is only one. From the definition of the inverse we can see that
A is the inverse of A−1 as well, since AA−1 = I and A−1 A = I.

Example 1.1. Let A be a nonsingular n × n matrix and let’s say we want to

solve n systems of equations that all have A as their coefficient matrix. Writing
the systems as matrix equations we would get

Ax = b1
Ax = b2
..
.
Ax = bn

Note that since A is nonsingular, all the solutions to these equations will be
unique. For notational purposes, let’s call them c’s. So we can rewrite the
equations replacing the x’s with c’s.

1
 Ac1 = b1
Ac2 = b2
..
.
Acn = bn

To solve the first of these equations, we row reduce the augmented matrix, and
since A is nonsingular, we know we’ll get the unique solution c1 (note, we don’t
know what c1 is until we compute it, we’ve just given it a name in advance
because we know it’s unique).
h i h i
A b1 ∼ In c1

And of course we do the same thing to solve all the other equations.

h i h i
A b2 ∼ In c2
h i h i
A b3 ∼ In c3
..
.
h i h i
A bn ∼ In cn

Notice that in all these row reductions, the first n columns of the augmented
matrix are A, and they always row-reduce to In . In the first system, b1 row-
reduces to c1 , in the second one, b2 row-reduces to c2 , etc. We can eliminate a
lot of steps by doing all these row reductions at once, as follows
h i h i
A b1 b2 · · · bn ∼ In c1 c2 · · · cn

We can simplify this even further if we write the b’s and c’s as matrices
h i h i
B = b1 b2 · · · bn , C = c1 c2 · · · cn

Then we have
h i h i
A B ∼ In C

2
To summarize, if we have a bunch of systems with the same nonsingular coeffi-
cient matrix, we can solve them all at once by forming a matrix B with all the
constant vectors, and the matrix C above will contain all the solution vectors.

Another interesting fact that will be useful later is that the equations

Ac1 = b1
Ac2 = b2
..
.
Acn = bn

tell us that
h i h i
AC = Ac1 Ac2 · · · Acn = b1 b2 · · · bn = B

So if A and B are n × n matrices and

h i h i
A B ∼ In C

then AC = B.

Example 1.2. Let

       
1 −1 0 0 1 −1
A = 1 0 1 , b1 = 1 , b2 = 2 , b3 =  2 
       

2 3 −1 5 1 7

Solve the following systems all at once.

Ax = b1
Ax = b2
Ax = b3

We simply form a matrix from the constant vectors, augment A with it, and
row reduce:

3
    
h i 1 −1 0 0 1 −1 1 0 0 1 1 1
A b1 b2 b3 = 1 0 1 1 2 2  ∼ 0 1 0 1 0 2
   

2 3 −1 5 1 7 0 0 1 0 1 1

Then we simply read the solution vectors from the second partition:

The solution set to Ax = b1 is  

n 1 o
1
 

0
The solution set to Ax = b2 is  
n 1 o
0
 

1
The solution set to Ax = b3 is  
n 1 o
2
 

Theorem 1.2. Computing a Matrix Inverse

If A is a nonsingular n × n matrix and D is the matrix formed by augmenting
A with In , then the reduced row-echelon form of D will be In augmented with
the n × n matrix C, where AC = In . We can write this as:
h i h i
If D = A In ∼ In C , then AC = In , and C = A−1

Proof. As we saw in example 1.1, if A and B are n × n matrices and

h i h i
A B ∼ In C

then AC = B. In this case, B = In , which is n × n, so we have the desired

result AC = In


4
Example 1.3. Let A be the matrix given below. Is A invertible? If it is, find
the inverse of A.
 
2 1 0
A = 0 3 2
 

1 1 2

We can answer both parts at once by augmenting A with I3 and row-reducing.

If A row-reduces to I3 , then A is invertible, and the second partition is A−1 .

  2 1 1

1 0 0 5 −5

2 1 0 1 0 0 5
  1 2 2

[A | I] = 0 3 2 0 1 0 ∼ 
0 1 0 5 −5

5
1 1 2 0 0 1 3
0 0 1 − 10 1
− 10 3
5

So A is invertible, and its inverse is:

 2

5 − 51 1
5
A−1 =
 
 1 2 2
 5 5 −5
3 1 3
− 10 − 10 5

Theorem 1.3. If A is an invertible n × n matrix, then the equation Ax = b

has the unique solution x = A−1 b for all b in Cn .

Proof. A−1 b is a solution to Ax = b since

AA−1 b = Ib = b

A−1 b is the unique solution, since if y is another solution, then

Ay = b
A−1 Ay = A−1 b
y = A−1 b

5
Theorem 1.4. Properties of Inverse Matrices

1. If A is invertible, then A−1 is invertible and (A−1 )−1 = A

2. If A and B are n×n and invertible, then so is AB, and (AB)−1 = B −1 A−1

3. If A is invertible, then so is AT , and (AT )−1 = (A−1 )T = A−T

4. If A is invertible and α is a non-zero scalar, then αA is invertible, and

(αA)−1 = 1 −1
αA

Theorem 1.5. If A and B are n × n matrices, then AB is nonsingular iff A

and B are both nonsingular

Theorem 1.6. If A and B are n × n matrices and AB = In , then BA = In .

Theorem 1.7. Equivalence of Nonsingular and Invertible

If A is an n × n matrix, then A is nonsingular iff A is invertible.

6
2 Invertible Matrix Theorem

Theorem 2.1. Invertible Matrix Theorem (IMT)

Let A be an n × n matrix. The following statements are logically equivalent.

1. A is an invertible matrix.

2. A is nonsingular

3. A is row equivalent to the n × n identity matrix.

4. A has n pivot positions.

5. The equation Ax = 0 has only the trivial solution.

6. N (A) = {0}

7. The columns of A form a linearly independent set.

8. The linear transformation x 7→ Ax is one-to-one.

9. The equation Ax = b has a unique solution for every b in Cn .

10. The system LS(A, b) has a unique solution for every b in Cn .

11. The columns of A span Cn .

12. The linear transformation x 7→ Ax is onto.

13. There is an n × n matrix C such that CA = I.

14. There is an n × n matrix D such that AD = I.

15. AT is an invertible matrix.

16. det A 6= 0.

7
Definition 2.1. Unitary Matrix
A unitary matrix U is an n × n matrix that satisfies the equation U ∗ U = In .

Theorem 2.2. If U is a unitary matrix, then U is invertible and U −1 = U ∗

Proof. If U is unitary, then U ∗ U = In . Since In is nonsingular, both U and U ∗

are nonsingular, and therefore invertible. We also know that since U and U ∗
are square and U ∗ U = In , U U ∗ = In . By the definition of an invertible matrix,
U −1 = U ∗ . 

Theorem 2.3. The columns of an n × n matrix U form an orthonormal set iff

U is a unitary matrix.

u∗1
 
 ∗
 u2  ∗
Proof. Let U = [u1 u2 · · · un ], and U = 
 .. . So

 . 
u∗n

u∗1
 
 ∗
 u2 
U ∗U = 
 ..  [u1 u2 · · · un ]

 . 
u∗n

u∗1 u1 u∗1 u2 · · · u∗1 un

 
 ∗
 u2 u1 u∗2 u2 · · · u∗2 un 

=
 .. .. .. 
 . . . 
u∗n u1 u∗n u2 · · · u∗n un

= I, iff u∗i ui = hui , ui i = 1 ∀i, and u∗i uj = hui , uj i = 0 for i 6= j.

So {u1 , u2 , . . . , un } is an orthonormal set iff U ∗ U = I.



8
Theorem 2.4. If U is an n × n unitary matrix and x and y are in Cn , then

hU x, U yi = hx, yi and kU xk = kxk

Proof.

hU x, U yi = (U x)∗ U y
= x∗ U ∗ U y
= x∗ In y
= x∗ y
= hx, yi

p
kU xk = hU x, U xi
p
= hx, xi
= kxk