Diagonalization

Elementary Linear Algebra

R. Larsen et al. (6 Edition)
The Matrix Diagonalization Problem

This

  is an example of similarity transformation.
A → P−1AP
This is important because it preserves many properties of the matr
ix A. For example, if we let D = P−1AP, then A and D have the sa
me determinant since
det(D) = det(P−1AP) = det(P−1)det(A)det(P)

In general, any property that is preserved by a similarity

transformation is called a similarity invariant and is said to be
invariant under similarity.
Table 1 lists the most important similarity invariants.
2/53
 The Matrix Diagonalization Problem
Table 1 Similarity Invariants
Property Description

Determinant A and P−1AP have the same determinant.

Invertibility A is invertible if and only if P−1AP is invertible.

Rank A and P−1AP have the same rank.

Nullity A and P−1AP have the same nullity.
Trace A and P−1AP have the same trace.

Characteristic polynomial A and P−1AP have the same characteristic polynomial.

Eigenvalues A and P−1AP have the same eigenvalues.

If λ is an eigenvalue of A (and hence of P−1AP) then the

Eigenspace dimension eigenspace of A corresponding to λ and the eigenspace of P−1AP
corresponding to λ have the same dimension.
3/53
 Diagonalization

Definition 1:
If A and D are square matrices, then we say that D is similar
to A if there is an invertible matrix P such that D = P−1AP.


Recall:
Two square matrices A and D are called similar if there exists
an invertible matrix P such that D = P−1AP.

Elementary Linear Algebra: Section 7.2, p.435 4/53

 Diagonalization

Definition 2:

A square matrix A is called diagonalizable if it is similar to

some diagonal matrix; that is, if there exists an invertible
matrix P such that P-1AP=D is a diagonal matrix.
In this case the matrix P diagonalizes A.


Recall:

Matrices that are similar to diagonal matrices are called

diagonalizable.

Elementary Linear Algebra: Section 7.2, p.435 5/53

 THEOREMS

Theorem 7.2.1
An nn matrix A is diagonalizable if and only if
it has n linearly independent eigenvectors.
Proof:
() A is diagonaliz able
there exists an invertible P s.t. D  P 1 AP is diagonal
Let P  [ p1 p2  pn ] and D  diag (1 , 2 , , n )
1 0  0
 0 2  0
PD  [ p1 p2  pn ] 
   
 0 0  n 
 [1 p1 2 p2  n pn ]
Elementary Linear Algebra: Section 7.2, p.437 6/53
 AP  A[ p1 p2  pn ]  [ Ap1 Ap2  Apn ]

 AP  PD
 Api  i pi , i  1, 2, , n
(i.e. the column vector pi of P are eigenvecto rs of A)
P is invertible  p1 , p2 , , pn are linearly independent.
 A has n linearly independent eigenvectors.

() A has n linearly independent eigenvectors p1 , p2 ,  pn

with corresponding eigenvalues 1 , 2 , n
i.e. Api  i pi , i  1, 2, , n
Let P  [ p1 p2  pn ]

Elementary Linear Algebra: Section 7.2, p.438 7/53

 AP  A[ p1 p 2  pn ]
 [ Ap1 Ap2  Apn ]
 [1 p1 2 p2  n pn ]
1 0  0 
0   0 
 [ p1 p2  pn ] 2   PD
    
 
 0 0  n 
 p1 , p1 , , pn are linearly independent  P is invertible
 P 1 AP  D
 A is diagonalizable
Note: If n linearly independent eigenvectors do not exist,
then an nn matrix A is not diagonalizable.
Elementary Linear Algebra: Section 7.2, p.438 8/53
 THEOREMS

Theorem 7.2.2

(a) If λ1, λ2, …, λn are distinct eigenvalues of a matrix A, and if

p1, p2, …, pn are corresponding eigenvectors, then

{p1, p2, …, pn} is a linearly independent set.

(b) An n × n matrix with n distinct eigenvalues is diagonalizable.


Recall:
A n x n matrix with n linearly independent eigenvectors has n
distinct eigenvalues.

Elementary Linear Algebra: Section 7.2, p.437 9/53

 Diagonalizing an n × n Matrix

Steps for diagonalizing an nn square matrix:

Step 1: Find n linearly independent eigenvectors p1 , p2 , , pn

for A with corresponding eigenvalues 1 , 2 , , n
Step 2: Let P  [ p1 p2  pn ]
Step 3: Let P−1AP will be a diagonal matrix = D
1 0  0 
0   0 
P 1 AP  D   2 , where Ap   p , i  1, 2,  , n
     i i i

 
 0 0  n 
Note:
The order of the eigenvalues used to form P will determine
the order in which the eigenvalues appear on the main diagonal of D.
Elementary Linear Algebra: Section 7.2, p.439 10/53
 Example:

Ex 1: (A diagonalizable matrix)
 𝐴 = 7 2
[− 4 ] 1
Sol: Characteristic equation:

Eigenvalues:
Eigenvectors: if ,

Elementary Linear Algebra: Section 7.2, p.436 11/53

  
Inverse of P:

Check if :

Therefore, matrix A is diagonalizable because the diagonal

entries of D are the eigenvalues of the given matrix A.

Elementary Linear Algebra: Section 7.2, p.436 12/53

 Example:

Ex 2: (Finding a Matrix P That Diagonalizes a Matrix A)
1 3 0 
A  3 1 0 
 
 0 0  2 
Sol: Characteristic equation:
 1  3 0
I  A   3   1 0  (  4)(  2) 2  0
0 0 2
Eigenvalues : 1  4, 2  2, 3  2
1 
(1)  4  Eigenvecto r : p1  1 (See p.403 Ex.5)
0
Elementary Linear Algebra: Section 7.2, p.436 13/53
 1 0 
(2)  2  Eigenvector : p2   1, p3  0 (See p.403 Ex.5)
 0  1

There are three basis vectors in total, so the matrix P

1 1 0 4 0 0
P  [ p1 p2 p3 ]  1  1 0  P 1 AP  0  2 0 
0 0 1 0 0  2
  Check if AP = PD,

Therefore, the matrix P diagonalizes A.

Elementary Linear Algebra: Section 7.2, p.436 14/53

 Example:

Ex 3: (A matrix that is not diagonalizable)
Show that the following matrix is not diagonalizable.
1 2 
A 
 0 1 
Sol: Characteristic equation:
 1  2
I  A   (  1) 2  0
0  1
Eigenvalue : 1  1

 0  2  0 1  1
I  A  I  A    ~   Eigenvecto r : p1   
 0 0  0 0  0 
A does not have two (n = 2) linearly independent
eigenvectors, so A is not diagonalizable.
Elementary Linear Algebra: Section 7.2, p.439 15/53
 Example:
2/306

  A Matrix that is Not Diagonalizable
Show that the following matrix is not diagonalizable:

Solution The characteristic polynomial of A is

Elementary Linear Algebra: Section 7.2, p.439 16/53

Example:
2/306

  A Matrix that is Not Diagonalizable
Solution
So the characteristic equation is
(λ − 1)(λ − 2)2 = 0
and the distinct eigenvalues of A are λ = 1 and λ = 2. We leave
it to you to show that the bases for the eigenspaces are

Since A is 3 × 3 matrix and there are only two basis vectors in t

otal, A is not diagonalizable.

17/53
 Example:
2/306

  A Matrix that is Not Diagonalizable
Alternative Solution: Finding the dimensions of the eigenspaces
For this example, the eigenspace corresponding to λ = 1 is the sol
ution space of the system

Since the coefficient matrix has rank 2 (verify), the nullity of this
matrix is 1 by Theorem 4.8.2, and hence the eigenspace correspon
ding to λ = 1 is one-dimensional.

Elementary Linear Algebra: Section 7.2, p.439 18/53

 Example:
2/306

  A Matrix that is Not Diagonalizable
Alternative Solution: Finding the dimensions of the eigenspaces
The eigenspace corresponding to λ = 2 is the solution space of the
system

This coefficient matrix also has rank 2 and nullity 1 (verify), so th

e eigenspace corresponding to λ = 2 is also one-dimensional. Sinc
e the eigenspaces produce a total of two basis vectors, and since t
hree are needed, the matrix A is not diagonalizable.

Elementary Linear Algebra: Section 7.2, p.439 19/53

 Example:
3/307

  Recognizing Diagonalizability
We saw in Example 3 of the preceding topic that

has three distinct eigenvalues: λ = 4, λ = , and λ = .

Therefore, A is diagonalizable and

for some invertible matrix P. If needed, the matrix P can be found

using the method shown in previous examples.

Elementary Linear Algebra: Section 7.2, p.439 20/53

 
Ex 4: (Diagonalizing a matrix)
 1  1  1
A 1 3 1
 
 3 1  1
Find a matrix P such that P 1 AP is diagonal.

Sol: Characteristic equation:

 1 1 1
I  A   1   3  1  (  2)(  2)(  3)  0
3 1  1

Eigenvalues : 1  2, 2  2, 3  3

Elementary Linear Algebra: Section 7.2, p.440 21/53

 1  2  1 1 1  1 0 1
 1I  A   1  1  1 ~ 0 1 0
   
 3  1 3   0 0 0 
 x1   t   1  1
  x2    0   t  0   Eigenvecto r : p1   0 
 x3   t   1   1 
2  2
 3 1 1  1 0  14 
 2 I  A    1  5  1 ~ 0 1 14 
   
3  1  1 0 0 0 
 x1   14 t  1 1
  x2    14 t   1 
t
4 
 1   Eigenvector : p   1
 2  
 x3   t   4   4 
Elementary Linear Algebra: Section 7.2, p.440 22/53
 3  3  2 1 1  1 0 1 
 3 I  A   1 0  1 ~ 0 1  1
   
 3  1 4   0 0 0 
 x1   t   1  1
  x2    t   t  1   Eigenvector : p3   1 
 x3   t   1   1 
 1 1  1
Let P  [ p1 p2 p3 ]   0  1 1 
 1 4 1 
 2 0 0
 P 1 AP  0  2 0
0 0 3
Elementary Linear Algebra: Section 7.2, p.440 23/53
 
Theorem 7.1.2: Diagonalizability of Triangular Matrices

Since the eigenvalues of a triangular matrix are the entries on

its main diagonal, thus, a triangular matrix with distinct
entries on the main diagonal is diagonalizable.

Elementary Linear Algebra: Section 7.2, p.442 24/53

 Example:

Ex 5: Diagonalizability of Triangular Matrices
1  2 1 
A  0 0 1
 
0 0  3

Sol: Because A is a triangular matrix,

its eigenvalues are the main diagonal entries.
1  1, 2  0, 3  3
These three eigenvalues are distinct, so A is diagonalizable.

Elementary Linear Algebra: Section 7.2, p.443 25/53

 Example:
4/307

  Diagonalizability of Triangular Matrices
Determine whether a triangular matrix A is diagonalizable.

Because A is a triangular matrix, A is a diagonalizable matrix

with eigenvalues λ1 = −1, λ2 = 3, λ3 = 5, λ4 = −2.

Elementary Linear Algebra: Section 7.2, p.439 26/53

 
Theorem 7.2.3.
If k is a positive integer, λ is an eigenvalue of a matrix A, and x is
a corresponding eigenvector, then λk is an eigenvalue of Ak and x i
s a corresponding eigenvector.

Note that diagonalizability is not a requirement in Theorem 7.2.3.

Elementary Linear Algebra: Section 7.2, p.439 27/53

 Example:
5/307

  Eigenvalues and Eigenvectors of Matrix Powers
In Example 2/306, we found the eigenvalues and corresponding ei
genvectors of the matrix

Do the same for A7.

Solution We know from Example 2/306 that the eigenvalues of A

are λ = 1 and λ = 2, so the eigenvalues of A7 are λ = 17 = 1 and λ =
27 = 128. The eigenvectors p1 and p2 obtained in Example 1/305
corresponding to the eigenvalues λ = 1 and λ = 2 of A are also the
eigenvectors corresponding to the eigenvalues λ = 1 and λ = 128 o
f A7.

Elementary Linear Algebra: Section 7.2, p.439 28/53

 Computing Powers of a Matrix

  (1)

Formula (1) reveals that raising a diagonalizable matrix A to a

positive integer power has the effect of raising its eigenvalues
to that power.


Notes: k is a positive integer
Elementary Linear Algebra: Section 7.2, p.445 29/53
 Examples
6/308
  Powers of a Matrix
Use (1) to find A13, where

Solution We showed in previous example that the matrix A is

diagonalized by

and that

Elementary Linear Algebra: Section 7.2, p.445 30/53

 Examples
6/308
  Powers of a Matrix
Solution
Thus, it follows from (1) that

(2)

Remark:
With the method in the preceding example, most of the work is in
diagonalizing A. Once that work is done, it can be used to compute
any power of A. Thus, to compute A1000 we need only change the
exponents from 13 to 1000 in (2).

Elementary Linear Algebra: Section 7.2, p.445 31/53

 Symmetric Matrices and Orthogonal Diagonalization

Symmetric matrix: NOT INCLUDED TOPICS
A square matrix A is symmetric if it is equal to its
transpose:T
A A

Ex 1: (Symmetric matrices and nonsymetric
matrices)
 0 1  2
A 1 3 0  (symmetric)
 
  2 0 5 
4 3
B  (symmetric)
 3 1
3 2 1
C  1  4 0 (nonsymmetric)
 
1 0 5
Elementary Linear Algebra: Section 7.3, p.447 32/53
 THEOREMS

Theorem 7.7: (Eigenvalues of symmetric matrices)

If A is an nn symmetric matrix, then the following properties

are true.
(1) A is diagonalizable.
(2) All eigenvalues of A are real.
(3) If  is an eigenvalue of A with multiplicity k, then  has k
linearly independent eigenvectors. That is, the eigenspace
of  has dimension k.

Elementary Linear Algebra: Section 7.3, p.447 33/53

 
Ex 2:
Prove that a symmetric matrix is diagonalizable.
a c 
A
 c b

Pf: Characteristic equation:

 a c
I  A   2  (a  b)  ab  c 2  0
c  b
As a quadratic in , this polynomial has a discriminant of
(a  b) 2  4(ab  c 2 )  a 2  2ab  b 2  4ab  4c 2
 a 2  2ab  b 2  4c 2
 (a  b) 2  4c 2  0
Elementary Linear Algebra: Section 7.3, p.448 34/53
 (1) (a  b) 2  4c 2  0

 a  b, c  0
a 0 
A  is a matrix of diagonal.
0 a 

(2) (a  b) 2  4c 2  0

The characteristic polynomial of A has two distinct real

roots, which implies that A has two distinct real eigenvalues.
Thus, A is diagonalizable.

Elementary Linear Algebra: Section 7.3, p.448 35/53

 
Orthogonal matrix:
A square matrix P is called orthogonal if it is invertible and
P 1  P T


Ex 3: (Orthogonal matrices)
 0 1 0  1
(a) P    is orthogonal because P  P  
1 T
 .
  1 0 1 0 
3  4  3 4
5 0 5   5 0
5
(b) P   0 1 0  is orthogonal because P 1  P T   0 1 0 .
4 3   4 3
 0   0 
 5 5   5 5 

Elementary Linear Algebra: Section 7.3, p.449 36/53

 THEOREMS

Theorem 7.8: (Properties of orthogonal matrices)

An nn matrix P is orthogonal if and only if

its column vectors form an orthogonal set.

Elementary Linear Algebra: Section 7.3, p.450 37/53

 
Ex 4: (An orthogonal matrix)
 13 2
3
2
3

 2 
P 5 1
5
0
 2 4 5 
3 5 3 5 3 5

Sol: If P is a orthogonal matrix, then P 1  P T  PPT  I

 1 2 2   13 2 2  1 0 0
 32 3 3
 2 5 3 5
  I
PP   5
T 1
0  3 1 4

  0 1 0
5 5 3 5

 3 25 4 5  2
0 5   0 0 1 
 3 5 3 5  3 3 5 
 13   23   23 
Let p1   25 , p2   15 , p3   0 
 325   345   3 5 5 
Elementary Linear Algebra: Section 7.3, p.451 38/53
 produces
p1  p2  p1  p3  p2  p3  0
p1  p2  p3  1
{ p1 , p2 , p3} is an orthonorma l set.

Elementary Linear Algebra: Section 7.3, p.451 39/53