Diagonalization of a Matrix:

Definition of diagonalization of a matrix:

A matrix A is diagonalizable if there exists a nonsingular matrix P and a diagonal
matrix D such that
D = P −1 AP .

Example:

Let
− 4 − 6
A= .
3 5 

Then,
−1
2 0  −1 − 2 − 4 − 6−1 − 2
D = = = P −1 AP ,
0 −1
 1 1  
 3 5 
 1 1 

where
2 0 −1 − 2
D = , P = .
0 −1 1 1 


Theorem:
An n × n matrix A is diagonalizable if and only if it has n linearly
independent eigenvector.

[proof:]

⇒:

A is diagonalizable. Then, there exists a nonsingular matrix P and a

diagonal matrix
λ1 0  0
0 λ  0 
D= 2
,
   
 
0 0  λn 
such that

1
 D = P −1 AP ⇔ AP = PD
λ1 0  0
0 λ2  0 
⇔ A[ col 1 ( P ) col 2 ( P )  col n ( P )] = [ col 1 ( P ) col 2 ( P )  col n ( P )] 
   
 
0 0  λn 
.
Then,
Acol j ( P ) = λj col j ( P ), j =1,2,  , n.

That is,
col 1 ( P ), col 2 ( P ),  , col n ( P )
are eigenvectors associated with the eigenvalues λ1 , λ2 ,  , λn .
Since P is nonsingular, thus col 1 ( P ), col 2 ( P),  , col n ( P ) are linearly
independent.

⇐:
Let x1 , x 2 ,  , x n be n linearly independent eigenvectors of A associated
with the eigenvalues λ1 , λ2 ,  , λn . That is,
Ax j = λj x j , j =1,2,  , n.

Thus, let
P = [ x1 x2  x n ] (i.e., col j ( P ) = x j )

and
λ1 0  0
0 λ  0 
D= 2
.
   
 
0 0  λn 
Since Ax j = λj x j ,
λ1 0 0

0 λ2  0 
AP = A[ x1 x2  xn ] = [ x1 x2  xn ]  = PD .
   
 
0 0  λn 
Thus,
P −1 AP = P −1 PD = D ,
P −1 exists because x1 , x 2 ,  , x n are linearly independent and thus P is
nonsingular.

Important result:
An n × n matrix A is diagonalizable if all the roots of its characteristic

2
equation are real and distinct.

Example:

Let
− 4 − 6
A= .
3 5 

Find the nonsingular matrix P and the diagonal matrix D such that
D = P −1 AP
and find A n , n is any positive integer.

[solution:]

We need to find the eigenvalues and eigenvectors of A first. The characteristic

equation of A is
λ +4 6
det( λI − A) = = ( λ +1)( λ − 2 ) = 0 .
−3 λ −5

⇒ λ = −1 or 2 .
By the above important result, A is diagonalizable. Then,

1. As λ = 2 ,
−1
Ax = 2 x ⇔ ( 2 I − A) x = 0 ⇔ x = r  , r ∈ R.
1 
2. As λ = −1 ,
−2
Ax = −x ⇔ ( − I − A) x = 0 ⇔ x = t  , t ∈R.
1 
Thus,
−1 − 2
 1  and  1 
   
are two linearly independent eigenvectors of A.

Let
−1 − 2 2 0
P =  and D =  .
1 1  0 −1

Then, by the above theorem,
D = P −1 AP .

3
To find A n ,

2 n 0 
Dn =  n
= ( P −1 AP )( P −1 AP ) ( P −1 AP ) = P −1 A n P
0 ( −1) 
n times
Multiplied by P and P −1 on the both sides,
−1
−1 −1 −1 − 1 − 2 2 n 0  − 1 − 2 
PD P n
= PP A PP n
=A =
n

1 1   0 ( − 1) n   1 1 
=
[
− 2 n + 2 ⋅ ( − 1) n +1 ] − [2 n +1
+ 2 ⋅ ( − 1)
n +1
]
 2 + ( − 1) 2 n +1 + ( − 1)
n n +1 n +1


Note (very important):

If A is an n × n diagonalizable matrix, then there exists an nonsingular
matrix P such that
D = P −1 AP ,
where col 1 ( P ), col 2 ( P),  , col n ( P ) are n linearly independent
eigenvectors of A and the diagonal elements of the diagonal matrix D
are the eigenvalues of A associated with these eigenvectors.

Note:
For any n × n diagonalizable matrix A, D = P −1 AP , then
A k = PD k P −1 , k =1,2,
where

λ1k 0  0
 
0 λk2  0
D =
k
.
   
 
 0 0  λkn 

Example:

5 − 3
Is A =  diagonalizable?
3 −1

[solution:]

4
 λ −5 3
det( λI − A) = = ( λ − 2) = 0 .
2

−3 λ +1

Then, λ = 2, 2 .
As λ = 2,
1
( 2 I − A) x = 0 ⇔ x = t  , t ∈R.
1
1
Therefore, all the eigenvectors are spanned by   . There does not exist two linearly
1 
independent eigenvectors. By the previous theorem, A is not diagonalizable.

Note:
An n × n matrix may fail to be diagonalizable since
 Not all roots of its characteristic equation are real numbers.
 It does not have n linearly independent eigenvectors.

Note:
The set S j consisting of both all eigenvectors of an n × n matrix A
associated with eigenvalue λj and zero vector 0 is a subspace of R n . S j
is called the eigenspace associated with λj .