Theorem (The Diagonalisation Theorem)

Let A be an n × n matrix. Then A is diagonalisable if and only if A has n

linearly independent eigenvectors.

P −1 AP is a diagonal matrix D if and only if the columns of P are n linearly

independent eigenvectors of A and the diagonal entries of D are the
eigenvalues of A corresponding to the eigenvectors of A in the same order.

Dr Scott Morrison (ANU) MATH1014 Notes Second Semester 2015 1 / 12

Example 1
Find a matrix P that diagonalises the matrix
 
−1 0 1
A =  3 0 −3 .
 
1 0 −1

The characteristic polynomial is given by

 
−1 − λ 0 1
det(A − λI) = det  3 −λ −3  .
 
1 0 −1 − λ

= (−1 − λ)(−λ)(−1 − λ) + λ
= −λ2 (λ + 2).

Dr Scott Morrison (ANU) MATH1014 Notes Second Semester 2015 2 / 12

The eigenvalues of A are λ = 0 (of multiplicity 2) and λ = −2 (of
multiplicity 1).
The eigenspace E0 has a basis consisting of the vectors
   
0 1
p1 = 1 , p2 = 0
   
0 1

and the eigenspace E−2 has a basis consisting of the vector

 
−1
p3 =  3 
 
1

It is easy to check that these vectors are linearly independent.

Dr Scott Morrison (ANU) MATH1014 Notes Second Semester 2015 3 / 12

 So if we take
 
h i 0 1 −1
P = p1 p2 p3 = 1 0 3 
 
0 1 1

then P is invertible.
 
0 0 0
It is easy to check that AP = PD where D = 0 0 0 
 
0 0 −2
    
−1 0 1 0 1 −1 0 0 2
AP =  3 0 −3 1 0 3  = 0 0 −6
    
1 0 −1 0 1 1 0 0 −2
    
0 1 −1 0 0 0 0 0 2
PD = 1 0 3  0 0 0  = 0 0 −6 .
    
0 1 1 0 0 −2 0 0 −2

Dr Scott Morrison (ANU) MATH1014 Notes Second Semester 2015 4 / 12

Example 2
Can you find a matrix P that diagonalises the matrix
 
0 1 0
A = 0 0 1?
 
2 −5 4

The characteristic polynomial is given by

 
−λ 1 0
det(A − λI) = det  0 −λ 1 
 
2 −5 4 − λ

= (−λ) [−λ(4 − λ) + 5] − 1(−2)

= −λ3 + 4λ2 − 5λ + 2
= −(λ − 1)2 (λ − 2)

Dr Scott Morrison (ANU) MATH1014 Notes Second Semester 2015 5 / 12

This means that A has eigenvalues λ = 1 (of multiplicity 2) and λ = 2 (of
multiplicity 1).
The corresponding eigenspaces are
   
 1 
   1 
 
E1 = Span 1 , E2 = Span 2 .
   

 1   
  4 

Note that although λ = 1 has multiplicity 2, the corresponding eigenspace

has dimension 1. This means that we can only find 2 linearly independent
eigenvectors, and by the Diagonalisation Theorem A is not diagonalisable.

Dr Scott Morrison (ANU) MATH1014 Notes Second Semester 2015 6 / 12

Example 3
Consider the matrix  
2 −3 7
A = 0 5 1 .
 
0 0 1
Why is A diagonalisable?

Dr Scott Morrison (ANU) MATH1014 Notes Second Semester 2015 7 / 12

Example 3
Consider the matrix  
2 −3 7
A = 0 5 1 .
 
0 0 1
Why is A diagonalisable?

Since A is upper triangular, it’s easy to see that it has three distinct
eigenvalues: λ1 = 2, λ2 = 5 and λ3 = 1. Eigenvectors corresponding to
distinct eigenvalues are linearly independent, so A has three linearly
independent eigenvectors and is therefore diagonalisable.

Dr Scott Morrison (ANU) MATH1014 Notes Second Semester 2015 7 / 12

Example 3
Consider the matrix  
2 −3 7
A = 0 5 1 .
 
0 0 1
Why is A diagonalisable?

Since A is upper triangular, it’s easy to see that it has three distinct
eigenvalues: λ1 = 2, λ2 = 5 and λ3 = 1. Eigenvectors corresponding to
distinct eigenvalues are linearly independent, so A has three linearly
independent eigenvectors and is therefore diagonalisable.
Theorem
If A is an n × n matrix with n distinct eigenvalues, then A is diagonalisable.

Dr Scott Morrison (ANU) MATH1014 Notes Second Semester 2015 7 / 12

Example 4
Is the matrix  
4 0 0 0
 0 4 0 0 
A=
 
0 0 2 0

 
1 0 0 2
diagonalisable?

The eigenvalues are λ = 4 with multiplicity 2, and λ = 2 with multiplicity

2.

Dr Scott Morrison (ANU) MATH1014 Notes Second Semester 2015 8 / 12

The eigenspace E4 is found as follows:
 
0 0 0 0
 0 0 0 0
E4 = Nul 
 
0 0 −2 0 


1 0 0 −2
    

 0 2 


 1 0

= Span v1 =   , v2 =   ,
   

 0 0

 

0 1 

and has dimension 2.

Dr Scott Morrison (ANU) MATH1014 Notes Second Semester 2015 9 / 12

The eigenspace E2 is given by
 
2 0 0 0
 0 2 0 0 
E2 = Nul 
 
0 0 0 0

 
1 0 0 0
    

 0 0 


 0 0

= Span v3 =   , v4 =   ,
   

 1 0

 

0 1 

and has dimension 2.

Dr Scott Morrison (ANU) MATH1014 Notes Second Semester 2015 10 / 12

        

 0 2 0 0 

1 0 0 0 
{v1 , v2 , v3 , v4 } =   ,   ,   ,   is linearly independent.
       
0 0 1 0
 
 

0 1 0 1 
h i
This implies that P = v1 v2 v3 v4 is invertible and A = PDP −1
where
   
0 2 0 0 4 0 0 0
 1 0 0 0   0 4 0 0 
P=  and D =  .
   
 0 0 1 0   0 0 2 0 
0 1 0 1 0 0 0 2

Dr Scott Morrison (ANU) MATH1014 Notes Second Semester 2015 11 / 12

Theorem
Let A be an n × n matrix whose distinct eigenvalues are λ1 , λ2 , . . . , λp .
1 For 1 ≤ k ≤ p, the dimension of the eigenspace for λk is less than or
equal to its multiplicity.
2 The matrix A is diagonalisable if and only if the sum of the
dimensions of the distinct eigenspaces equals n.
3 If A is diagonalisable and Bk is a basis for the eigenspace
corresponding to λk for each k, then the total collection of vectors in
the sets B1 , B2 , . . . , Bp forms an eigenvector basis for Rn .
4 If P −1 AP = D for a diagonal matrix D, then P is the change of basis
matrix from eigenvector coordinates to standard coordinates.

Dr Scott Morrison (ANU) MATH1014 Notes Second Semester 2015 12 / 12