CHAPTER-1 EIGENVALUES AND EIGENVECTORS

CHAPTER 1
EIGENVALUES,EIGENVECTORS AND
DIAGONALIZATION
1.1 Eigenvalues and eigenvectors of matrix
1.2 Similarity of Matrices and Diagonalization
1.3 Polynomials and Matrices
1.4 Eigenvalues and Eigenvectors of a Linear Operator

Introduction
This chapter is concerned with the so-called "diagonalization problem." We are going to
investigate matrices and (or) linear operators that can be replaced by diagonal matrices.

We begin this investigation with the introduction of special scalars and vectors, called
eigenvalues and eigenvectors respectively that provide us with the tools necessary to
describe diagonalizable matrices.
Given a square matrix A , we seek to find an invertible matrix P such that P 1 AP is a
diagonal matrix. In particular an algorithm is given to find the matrix P when exists.

Given a linear operator T : V  V , where V is a finite-dimensional vector space, we seek

answers to the following questions:
1. Does there exist a basis  for V such that T  is a diagonal matrix?
2. If such a basis exists, how can it be found?

Objectives
At the end of the lesson, students will be able to:
 evaluate eigenvalues and eigenvectors of matrix or linear operators.
 identify similar matrices.
 identify whether a matrix or linear operators is diagonalizable or not.
 find an invertible matrix P and a diagonal matrix D such that P1 AP  D if
possible.
 find a basis  for V such that T  is a diagonal matrix if possible.

BDU,DEPARTEMENT OF MATHEMATICS | 3
 Math 2042 LINEAR ALGEBRA II

1.1 Eigenvalues and Eigenvectors

Definition 1.1.1 Let A   aij  be a square matrix with elements in  or  . A scalar
nn

 is called an eigenvalue of matrix A corresponding to a non -zero eigenvector v if

Av   v (1)

It is clear that the zero vector would satisfy equation (1) for any scalar  , so the zero vector
is not allowed to be an eigenvector.
Notes: Eignvectors are never equal zero!

0 1 0 1
   
Example 1.1.1 The vector v   1  is an eingenvector of A   1 2 0  corresponding to
0 3 0 1
   

0
 
the eigenvalue   2 , since Av   2   2v
0
 
Notes: Eigenvectors and eigenvalues come in pairs.

How to find eigenvalues and eigenvectors of a matrix?

Let A   aij  be an n  n matrix. Write Av   v as

nn

Av   I n v or equivalently ( I n  A)v  0 (2 (a))

In a matrix form
   a11  a12  a13   ann  v1   0 
  
 a21   a22  a23   a2 n  v2   0 
 (2 (b))
          
    
 an1  an 2  an3    ann  vn   0 

for  to be an eigenvalue, there must be a non-zero solution vector of the homogeneous

system equation (2). Equation (2) has a non zero solution if and only if
det( I n  A)  0 (3)

4 | BDU,DEPARTEMENT OF MATHEMATICS
 CHAPTER-1 EIGENVALUES AND EIGENVECTORS

    a11  a12  a13   ann  

 
a21   a22  a23   a2 n  
In other words det   0
     
  
  an1  an 2  an 3    ann  

The matrix ( I n  A) is called characteristic matrix of A .

p( )  det( I n  A) is called a characteristic polynomial of A .

The equation p( )  det( I n  A)  0 is called characteristic equation of A .

Solving equation (3) gives the eigenvalues of A .

Remark: In the above discussion ( I n  A)v  0 equivalent to ( A   I n )v  0 . Sometimes
we will use ( A   I n )v  0 . It makes absolutely no difference and you should use which ever
you like better. (See exerscise 1.1 problem 2(b))
 3 3
Example 1.1.2 Find the eigenvalues and corresponding engienvectors of A   
1 5
Solution. To find the eigenvalues of A , find the roots of characteristic polynomial.
   3 3 
The characteristic matrix of A ,   I 2  A   
 1   5 
    3 3  
det( I n  A)  0 , that is, det     0
  1   5  
In other words  2  8  12  0 that is (  2)(  6)  0
Thus 1  2 and 2  6 are eigenvalues of A .

An engienvector of A corresponding to the eigenvalue 1  2 is the solution of the

homogenuou system (2I 2  A)v  0 , that is,

 1 3  x   0 
     
 1 3  y   0 
 3t 
All solutions of the homogeneous system have the form   , where t   .
 t 
 3 
v1    is an engienvector of A corresponding to the eigenvalue 1  2 .
1
Remember that you can't take t  0 because this would result in the zero vector.

BDU,DEPARTEMENT OF MATHEMATICS | 5
 Math 2042 LINEAR ALGEBRA II

  3  
Thus v  t   : t    0 is the set of all engienvectors of A corresponding to 1  2 .
 1 
Similarly an eigenvectors of A corresponding to 2  6 is the solution of the system

(6I 2  A)v  0 .

 3 3  x   0 
     
 1 1  y   0 
t 
solutions of the homogeneous system have the form   , where t   .
t 
 1
v2    is an engienvector of A corresponding to the eigenvalue 2  6 .
 1
Again remember that you can't take t  0 because this would result in the zero vector.
  1 
Thus v  t   : t    0 is the set of engienvectors of A corresponding to 2  6 .
  1 

Definition 1.1.2 Let A   aij  and  is an eigenvalue of A . The eigenspace of A

nn

for  , denoted by     , is the set of all the eigenvectors

A of A corresponding to ,

together with the inclusion of the zero vector, that is,

     v  F
A
n
: ( I n  A)v  0 , where F is the field  or  .

The eigenspace of A for  is the nullspace of the matrix  I n  A .

Thus the eigenspace of A is the kernel of   I n  A denoted by ker   I n  A .

Theorem 1.1.1 Suppose A   aij  and  is an eigenvalue of A . Then the eigenspace

nn

of A ,     , is a subspace of the vector space F

A
n
.

Proof. 1) By definition the zero vector is in     , thus     set is non empty.

A A

2) Suppose u, v      , that is, u and v are two eigenvalues of

A A for 

6 | BDU,DEPARTEMENT OF MATHEMATICS
 CHAPTER-1 EIGENVALUES AND EIGENVECTORS

then A  u  v   Au  Av  u  v    u  v  So either u  v  0 or

u  v is an eingenvector of A corresponding to .
3) Suppose a scalar c  F and u      , that is, u is an eigenvalue of A for  then
A

A(cu)  c( Au)  c(u)   (cu) . So either cu  0 or cu is an eigenvector.

Example 1.1.3 Find the eigenvalues and corresponding eingienvectors of the matrix
 2 2 2 
 
A   1 3 1 
 1 1 1 
 

    2 2 2 
 
Solution. det   I3  A  0 in other words det   1   3 1     3  6 2  8  0
 1 1   1 

Hence 1  0, 2  2 and 3  4 are eigenvalues of A .
Notes: 0 can be an eigenvalue!
An engienvector of A corresponding to the eigenvalue 1  0 is the solution of the system

(0I3  A)v  0 that is

 2 2 2 
 
 1 3 1  v  0
 1 1 1
 

 1 0 1
the reduced row echelon form is  0 1 0 
0 0 0 
 

 1 
   
Hence v1  t  0  : t    is the eignspace of A corresponding to 1  0 .
 1 
   

 1 
   
Observe that t  0  : t    is a line through the origin which is asubspace of  3 .
 1 
   
Similarly for 2  2 is the solution of the system (2I3  A)v  0 , that is,

BDU,DEPARTEMENT OF MATHEMATICS | 7
 Math 2042 LINEAR ALGEBRA II

 0 2 2 
 
 1 1 1  v  0
 1 1 1 
 

1 0 0 
the reduced row echelon form  0 1 1
0 0 0 
 

 0 
   
Hence v2  t  1  : t    is eignspace of A corresponding to =2.
 1 
   

For 3  4 is the solution of the system (4I3  A)v  0 .

 2 2 2   1 1 0 
   
 1 1 1  the reduced row echelon form 0 0 1
 1 1 3  0 0 0
   

 1 
   
Hence v3  t  1  : t    is the eignspace of A corresponding to =4.
 0 
   

Theorem 1.1.2 If A   aij  is a triangular or diagonal matrix then the eigenvalues of

nn

A are the diagonal elements of A .

The proof is left as an exercise. (Hint: Recall the determinate of triangular and diagonal
matrices)

Let A   aij  matrix then det   I n  A  pA ( ) is a polynomial of degree n . By the

nn

fundamental theorem of algebra, we know that the characteristic polynomial pA ( ) can be

written as the product of linear factors over the complex numbers in the form

pA ( )     1     2     k  k .
m1 m2 m

where 1 , 2 ,, k are the distinct eigenvalues of A and the positive integers m1 , m2 ,, mk
are their respective multiplicities as the zeros of the characteristic polynomial.

8 | BDU,DEPARTEMENT OF MATHEMATICS
 CHAPTER-1 EIGENVALUES AND EIGENVECTORS

Definition 1.1.3 Let A   aij  with pA ( )     1     2     k  k ,

m1 m2 m
nn

where 1 , 2 ,, k are the distinct eigenvalues of A and m1 , m2 ,, mk are the positive

integers. The algebraic multiplicity of i is mi . The geometric multiplicity of i is

the dimension of the eigenspace of i , that is,dim( ker  i I n  A ).

Example 1.1.4 Find the algebraic and geometric multiplicities of eigenvalues of

3 1 1
 
A  0 3 2
0 0 1
 
Solution. det   I3  A  (  3)2 (  1)  0 , by theorem 1.1.2

1  3 has algebraic multiplicity m1  2 and 2  1 has algebraic multiplicity m2  1 .

To find the geometric multiplicities we need to find the corresponding eigenspaces.
We start with 1  3
 0 1 1  0 1 0
   
3I 3  A   0 0 2  whose reduced row echelon form 0 0 1
0 0 2  0 0 0
   
 1 
   
t  0  : t    is the eignspace of A corresponding to 1  3 .
 0 
   
The dimension of the eignspace is 1.
Hence the geometric multiplicity for the eigenvalue 1  3 is 1 .
Verify the geometric multiplicity for the eigenvalue 2  1 is also 1 .
Warning: In the above examples we use row operations after eigenvalues are determined
algebraically. You may have noticed it is a bit long. Sometimes students try to first row
reduce the matrix before looking for eigenvalues. This is a terrible idea because row
operations destroy the eigenvalues. The eigenvalue problem is really not about row
operations.

BDU,DEPARTEMENT OF MATHEMATICS | 9
 Math 2042 LINEAR ALGEBRA II

Example 1.1.5 Complex Eigenvalues

Find the eigenvalues and corresponding eingienvectors of
1 0 0 
 
A   0 2 1
0 1 2 
 
Solution. det   I3  A  ( 2  4  5)(  1)  0

The solutions are 1  1, 2  2  i and 3  2  i

There is nothing new about finding the eigenvectors for 1  1 .

Eigenvectors for 2  2  i
1
R  R
1  i 0 0  1 1  i 1  1 0 0  1 0 0 
     
 2  i  I3  A   0 i 1  R2  iR2  0 1 i   0 1 i 
 0 1 i   0 1 i  R  R  R  0 0 0 
    3 3 2  
0
 
Hence eigenvectors corresponding to 2  2  i is of the form t  i  , t  0
1
 
Eigenvectors for 2  2  i
1
R  R
1  i 0 0  1 1  i 1  1 0 0  1 0 0
     
 2  i  I3  A   0 i 1  R2  iR2  0 1 i  0 1 i 
 0 1 i   0 1 i  R  R  R  0 0 0 
    3 3 2  
0
 
Thus eigenvectors corresponding to 3  2  i is of the form t  i  , t  0
1
 
Example 1.1.6 Eigenvalues and Eingienvectors of Complex Matrix
Find the eigenvalues and corresponding eingienvectors of
 i i 1 
 
A   1 1 i 
 0 0 1 i
 
Solution. det   I3  A   (  (1  i))2  0
1  0 and 2  1  i are eigenvalues of A .

10 | BDU,DEPARTEMENT OF MATHEMATICS
 CHAPTER-1 EIGENVALUES AND EIGENVECTORS

 i i 1  R1  iR1  1 1 i 
   
For 1  0 , Av  0 A   1 1 i   1 1 i 
 0 0 1 i  0 0 1 i
   

 1 1 i   1 1 i 
   
R2  R2  R1  0 0 0  0 0 1 i
0 0 1 i R  R 0 0 0 
  2 3

1
 
Hence eigenvectors corresponding to 1  0 is of the form t  1  , t  0 .
0
 
For 2  1  i , 1  i  I 3  A v  0
 1 i 1  1 i 1 
 1  i  I3  A   1 i i  R2  R2  R1  0 0 1  i 
 
0 0 0  0 0 0 
   
 i 
 
Hence eigenvectors corresponding to 2  1  i is of the form t  1  , t  0 .
0
 
EXERCISES 1.1_________________________________________________
1. Determine the eigenvalues and eigenvectors of A.
 1 6 12   2 1 0  3 3 2 
     
a) A   0 13 30  b) A   1 0 0 c) A   1 1 2 
 0 9 20  0 0 3   1 3 4 
    
1 1 0 0  8 5 6 0
1 5 4    
  1 1 0 0 0 2 0 0
d) A   0 2 6  e) A   f) A  
0 0 3 0 0 1 1  10 5 8 0 
     
0 0 1 1  2 1 1 2
 4 2 2   4 2 0  1 1 6 
     
g) A   0 2 2  h) A   2 4 0  i) A   7 5 6 
2 0 2   2 2 2   1 2 
     7
1 0 0  i 0 1 
 i 1    
j) A    k) A   0 i 1  i  l) A   0 i 2  i 
2 i 1 1 i i  0 2  i i 
 
2. a) Prove theorem 1.1.2
b) If A is an n  n matrix. Show that det( I n  A)   1 det( A   I n )
n

BDU,DEPARTEMENT OF MATHEMATICS | 11
 Math 2042 LINEAR ALGEBRA II

3. Find the eigenvalues of the following matrix by inspection.

 1 
 3 0 0 0   1 3 7 11
   
 3 0 0  0 1 0 0 1
   0 3 8
a) A   2 7 0  b) A   3  c) A   2 
  
 4 8 0
   0 0 1 0 0 0 0 4
 
 1  0 0 0 2 
 0 0 0 
 2
4. Show that the standard basis vectors e1 , e2 ,, en for  n are eigenvectors for any
n  n diagonal matrix. What are the associated eigenvalues?
5. Find the algebraic and geometric multiplicities of eigenvalues of the following
Matrices.
3 0 1 1 1 0
 3 1    
a) A    b) A   0 3 2  c) A   0 3 1 
 1 3  0 0 1  1 1 2 
   
6. Show that
a) if  is an eingenvalue of A then the matrix (  I n  A ) is non-invertible.
b) A   aij  is invertible if and only if =0 is not an eigenvalue of A .
nn

7. Let A   aij  and pA ( )  det( I n  A)     1     2     k 

m1 m2 mk
the
nn

characteristic polynomial of A . Show that

a) det  A  1m1 2m2 kmk
b) det  A  (1)n a0 , where a0 is the constant term of det( I n  A)

c) tr  A  i 1 i , where 1 , 2 ,, n are the n eigenvalues of A not

n

necessarily distinct and tr  A is the sum of the diagonal elements

of A , that is, tr  A  i 1 aii .

n

8. Let A   aij  be a square matrix with real entries, and let  be the complex
nn

eigenvalue of A such that Av   v . Prove that

a) the complex conjugate of  is an eigenvalue of A .
b) the complex conjugate of v is an eigenvector of A corresponding to  .
___________________________________________________________

12 | BDU,DEPARTEMENT OF MATHEMATICS
 CHAPTER-1 EIGENVALUES AND EIGENVECTORS

Properties of eigenvalues and eigenvectors of a matrix

Theorem 1.1.3 Let A   aij  be a square matrix then

nn

1) If  is an eigenvalue of A , then  is an eigenvalue of  A , where  is a scalar.

2) If u and v are both eigenvectors of A for the same eigenvalue , then u  v is an
eigenvector of A for  or u  v  0
3) If  is an eigenvalue of A , then m is an eigenvalue of Am , where m is
non-negative integer.
1
4) If A is invatiable and  is an eigenvalue of A then is an eigenvalue of A1 .

5) If  is an eigenvalue of A , then  is an eigenvalue of At .

6) If S  v1 , v2 ,, vn  is the set of eigenvectors corresponding to eigenvalues

1 , 2 ,,.n such that i   j , whenever i  j .
Then S is linearly independent set.

Proof.
1) Let v  0 be an eigenvector of A for  then ( A)v   ( Av)   (v)  ( )v
3) The proof is by induction. Let v  0 be eigenvector of A corresponding to  .
For m  0 , we have Am v  A0 v  I n v  v   0 v

Assume the theorem is true for k  0 , that is, Ak v   k v

Ak 1v  A k ( Av) since matrix multiplication is associative.

 Ak   v  since v is an eigenvector of A corresponding to 

   Ak v  property of scalar multiple

    k v  by induction assumption.

  k 1v
In the above proof, observe that if v is an eigenvector of A with corresponding
eigenvalue  then v eigenvector of Am with corresponding eigenvalue  m .
1
4) does not involve division by zero, since A is invertible. (See exerscise 1.1problem

6(b)). Let v  0 be eigenvector of A corresponding to  . Then

BDU,DEPARTEMENT OF MATHEMATICS | 13
Math 2042 LINEAR ALGEBRA II

1  1
A1v  A1  ( v)    A1 ( v) 
  
1
 A1 ( Av) since v is an eigenvector of A corresponding to 


1

 A A v since matrix multiplication is associative.
1

1 1
 In v  v
 
In the above proof, observe that if v is an eigenvector of A with corresponding
1
eigenvalue  then v is an eigenvector of A1 with corresponding eigenvalue .

5) It suffices to show A and At have the same characteristic polynomial. Then their
eigenvalues
are identical.


PA ( )  det   I n  A  det   I n  A  since det  B   det  B  , for any matrix B .
t t

 det    I   A   det    I   A   P   
t t t
n n At

6) The proof is by induction

If n  1, the set S  v1  is linearly independent, since v1  0 . Suppose S is linearly

dependant. Define Si  v1 , v2 ,, vi  , where 1  i  n

Let Sk 1  v1 , v2 ,, vk 1  be linearly independent and Sk  v1 , v2 ,, vk 

be linearly dependent , where 2  k  n . Since v1 , v2 , , vk  is linearly dependant

there are scalars a1 , a2 , , ak not all zero such that

0  a1v1  a2 v2    ak vk (1)

We have 0   k I n  A 0

  k I n _ A a1v1  a2 v2    ak vk 
  k I n  A av1   k I n  A a2 v2     k I n  A  ak vk

 a1  k I n  A  v1  a2  k I n  A  v2    ak  k I n  A  vk
 a1  k I n v1  Av1   a2  k I n v2  Av2     ak  k I n vk  Avk 
 a1  k v1  1v1 1  a2  k v2  2 v2     ak  k vk  k vk  ,
but k vk  k vk  0

14 | BDU,DEPARTEMENT OF MATHEMATICS
 CHAPTER-1 EIGENVALUES AND EIGENVECTORS

Thus a1  k  1  v1  a2  k  2  v2  ...  ak 1 (k  k 1 )vk 1  0

This is a relation on the linear independent set v1 , v2 ,, vk 1  so the scalars all must

be zero. That is ai  i  k   0 , for 1  i  k  1 . Since i   j for i  j

Thus ai  0 , for 1  i  k  1 .

From (1) v1 , v2 , vk  linear independence with the simple equation ak vk  0

implies ak  0 since vk  0 . Thus S is linear independent set.

Example 1.1.7 Let A and B be n  n matrices. Show that AB and BA have the same
eigenvalues.
Solution. i) If 0 is an eigenvalue of AB then AB is singular. It follows one of A or B is
singular. Thus BA is singular hence 0 is an eigenvalue of BA .
ii) If   0 is eigenvalue of AB . Then there exists a non zero vector v such that
ABv   v . Set Bv  u , since   0 and v  0
ABv  A  Bv  =Au   v  0 . Hence u  0

Now BAu  BA( Bv)  B( ABv)  B(v)   ( Bv)  u

u is an eigenvector of BA belonging to the eigenvalue 
Thus AB and BA have the same eigenvalues.

EXERISES 1.2_________________________________________________
1. Find the eigenvalue of A9 , where

 1 3 7 11
   1 2 2 
0 1 3 8   
a) A=  2  b) A   1 2 1 
0 0 0 4   1 1 0 
   
0 0 0 2 
2. Express the eigenvalues of the following matrices in terms of eigenvalues of A .
t 1
1  1 
i) 2A ii) A 3
iii)  A iv)  A  v) A  3I n
3  2 

BDU,DEPARTEMENT OF MATHEMATICS | 15
 Math 2042 LINEAR ALGEBRA II

3. Let A be a 3  3 matrix whose characteristic polynomial PA         1   2 

What is the characteristic polynomial of A2 ? Is A2 invertible? Why?

4. A square matrix A is called idempotent if A2  A . Prove that if A is idempotent then
  0 or   1 are the only eigenvalue of A .
5. A square matrix A is called nilpotent if there exists a positive integer k such that
Ak  0 . Prove that if A is nilpotent 0 is the only eigenvalue of A .

6. Let A   aij  and Am  A for some positive integer m greater than 1. Show that if 
nn

is an eigenvalue of A then   0 or   1 .

7. If A and B are n  n matrices and v is an eigenvector of both A and B ,

then v is an eigenvector of AB .
8. Let A and B are two n  n invertible matrices which commute, that is, AB  BA
Suppose v is an eigenvector of B show that Av must be an eigenvector for B .
___________________________________________________________

1.2 Similarity of Matrices and Diagonalization

The notion of matrices being ”similar” is a lot like saying two matrices are row-equivalent.
Two similar matrices are not equal, but they share many important properties. In this section
we will discover what these common properties are.

Definition 1.2.1 Let A and B be two n  n matrices. Then B is similar to A , denoted by

B  A if there exits a non singular matrix P of n  n such that B  P1 AP .

Theorem 1.2.1 Similarity of matrices is an equivalence relation.

More precisely if A, B and C are matrices of size n  n . Then
1) A is similar to A . (Reflexive)
2) If A is similar to B , then B is similar to A . (Symmetric)
3) If A is similar to B and B is similar to C , then A is similar to C . (Transitive)

16 | BDU,DEPARTEMENT OF MATHEMATICS
 CHAPTER-1 EIGENVALUES AND EIGENVECTORS

Proof.
1) Every square matrix A is similar to itself, because A  I n1 AI n  I n AI n  A
1
2) A  B there exists invertible matrix P such that A  P BP .

It follows B   P 1  AP 1 . Hence B  A .
1

3) Since A  B and B  C , there exists invertible matrices P and R such that

A  P1BP and B  R1CR since P and R are invertible it follows RP is

 
invertible and  RP  C ( RP)  P 1 R 1 C ( RP)  P 1 ( R 1CR) P  P 1 BP  A
1

So A  ( RP)1 C ( RP) . Thus A  C .

Theorem 1.2.2 Similar matrices have equal eigenvalues.

If A and B are similar matrices then the characteristic polynomials of A and B are
equal, that is, PA     PB    .

Proof. Let A and B be n  n similar matrices. Then A  P1 BP

PA ( )  det   I n  A  det   I n  P 1 BP   det  P 1 I n P  P 1 BP 

 det  P1 ( I n  B) P   det  P 1 ) det( I n  B) det( P)   det  B   I n   PB ( )

Since det  P 1  det  P   1

Remark: The converse of the above theorem is not true.

 1 1 1 0
Example 1.2.1 Consider A    and B    and PA     1     PB ( )
2

 0 1 0 1
If the converse is true, that is A  B . Then there is an invertible matrix P such that
A  P1 BP . But A  P1 BP  P1 I 2 P  P1 P  I 2 .

But A  I 2 this contradiction tells us the converse if false.

BDU,DEPARTEMENT OF MATHEMATICS | 17
 Math 2042 LINEAR ALGEBRA II

EXERCISES 1.3_____________________________________________
1. Show that
a) the only matrix similar to the I n is I n itself.
b) The only matrix similar to the zero matrix is the zero matrix itself.
2. a) Define diagonal matrix.
b) Show that the product of two diagonal matrices is a diagonal matrix.
 1 0 0  0
 
0 2 0  0
c) If A   is a diagonal matrix then show that
    
 
0 0 0  n 

 1m 0 0  0
 
0  m
0  0
A 
m 2  , for any positive integer m .
    
 m 

 0 0 0  n1 
3. Let A and B be two similar matrix show that
a) trA  trB
b) At and B t are similar matrix.
c) A is invertible if and only if B is invertible.
d) A and B have the same rank.
e) A and B have the same nullity.
________________________________________________________________

Diagonalization
Good things happen when a matrix is similar to a diagonal matrix. For example, the
eigenvalues of the diagonal matrix are the diagonal entries. It can be a much simpler matter
to compute high powers of the matrix. Diagonalizable matrices are interesting in more
abstract settings.

Definition 1.2.2 A square matrix A is diagonalizable if it is similar to a diagonal matrix.

18 | BDU,DEPARTEMENT OF MATHEMATICS
 CHAPTER-1 EIGENVALUES AND EIGENVECTORS

In other words A is a diagonalizable if there is an invertible matrix P such that P1 AP  D

is a diagonal matrix. The matrix P is called a diagonalizing matrix for A.
Equivalently AP  PD , where P is an invertible
.
Theorem 1.2.3 An n  n matrix A is diagonalizable if and only if A has n linearly
independent eigenvectors.

Proof.
(  ) since A is assumed to be diagonalizable there is an invertible matrix
 p11 p12  p1n 
 
p p22  p2 n 
P   21 such that P1 AP  D , where D a diagonal matrix.
    
 
 pn1 pn 2  pnn 

 1 0 0  0
 
0 2 0  0
D It follows AP  PD , that is,
    
 
0 0 0  n 

 p11 p12  p1n  1 0 0  0

  
p p22  p2 n  0 2 0  0
Ap   21
         
  
 pn1 pn 2  pnn  0 0 0  n 

 1 p11 2 p12  n p1n 


p 2 p22  n p2 n 
  1 21 (1)
    
 
 1 pn1 2 pn 2  n pnn 

If we let P1 , P2 ,, Pn denotes the column vectors of P

AP1  1 P1 , AP2  2 P2 ,, APn  n Pn (2)

Since P is invertible, its columns all non zero. Thus it follows that 1, , 2 ,, n are

eignevalues of A and P1 , P2 ,, Pn are corresponding eigenvectors.

Since P is invertible, it follows that P1 , P2 ,, Pn are linearly independent.

Thus A has n linearly independent eigenvectors.

BDU,DEPARTEMENT OF MATHEMATICS | 19
Math 2042 LINEAR ALGEBRA II

  Assume that A has n linearly independent eigenvectors P1 , P2 ,, Pn with

corresponding eigenvalues. 1 , 2 ,, n and let

 p11 p12  p1n 

 
p p22  p2 n 
p   21 be the matrix whose column vectors are P1 , P2 ,, Pn
    
 
 pn1 pn 2  pnn 

The column vectors of the product AP are AP1 , AP2 ,, APn .

But Ap1  1 p1 , Ap2  2 p2 ,, Apn  n pn .

 1 p11 2 p12  n p1n 


p 2 p22  n p2 n 
Ap   1 21
    
 
 1 pn1 2 pn 2  n pnn 

 p11 p12  p1n  1 0 0  0

  
p p22  p2 n  0 2 0  0
  21
         
  
 pn1 pn 2  pnn  0 0 0  n 

= PD , where D is a diagonal matrix having eigenvalues 1 , 2 ,, n .

Since column vectors of P are linearly independent P is invertible.
Thus P1 AP  D , that is, A is diagonalizable.
Diagonalization Algorithm. Let A be an n  n matrix
1. Find the characteristic polynomial pA ( x) of A .

2. Find the roots of pA ( x) to obtain the eigenvalues of A .

3. For each eigenvalues  of A find a basis for the solution of   I n  A v  0 .

(These basis vectors are linearly independent eigenvectors of A belonging to  )

4. Consider the collection of S  v1 , v2 ,, vm  all eigenvectors obtained in step 3.

a) If m  n , then A is not diagonalizable.

b) If m  n , then A is diagonalizable. Specifically, set P be the matrix whose
columns are the eigenvectors v1 , v2 ,, vm . Then D  P1 AP  diag (1 , 2 ,, n ) ,

where i is an eigenvalue corresponding to the eigenvector vi .

20 | BDU,DEPARTEMENT OF MATHEMATICS
 CHAPTER-1 EIGENVALUES AND EIGENVECTORS

Example 1.2.2 Show that the following matrices are diagonalizable. Find the invertible
matrix P and the diagonal matrix D .
 1 1 1  0 0 2 
2 1    
a) A  b) A   1 3 1  c) A   1 2 1 
1 2  3 1 1 1 0 3 
   

Solution. a) det   I 2  A  (  2)2  1  0 Implies   1or   3

 1
For 1  1 the eigenspace is spanned by v1    and
1
 1
For 2  3 the eigenspace is spanned by v2   
 1
 1 1 1 0
Thus P    and D  
1
 . Verify that P AP  D
 1 1  0 3
Solution. b) P( )  det   I3  A  0 , that is,  3  3 2  4  12  (  2)(  2)(  3)  0
1  1  1
     
For 1  2 , we have v1   1 , for 2  2 , v2   0  , and for 3  3 , v3   1 
4 1 1
     
 1 1 1  2 0 0 
   
Thus P   1 0 1  and D   0 2 0  . Verify that P1 AP  D
 4 1 1   0 0 3
   
Solution. c) P( )  det   I3  A  (  2)2 (  1)  0 . Thus 1  1, 2  2 and 3  2 are
eigenvalues of A . Observe that the eigenvalues of A are not distinct.
 2 
 
For   1 the eigenspace is spanned by v1   1 
1
 
 1  0
   
For   2 the eigenspace is spanned by v2   0  and v3   1  . Verify.
1  0
   
Hence
 1 0 2   2 0 0
   
P   0 1 1  and D   0 2 0  .
1 0 1 0 0 1
   

BDU,DEPARTEMENT OF MATHEMATICS | 21
 Math 2042 LINEAR ALGEBRA II

Example 1.2.3 Matrix that is not diagonalizable.

 1 0 0
 
A   1 2 0
 3 5 2 
 
Solution. P( )  det   I3  A  (  2)2 (  1)  0 . Thus 1  1, 2  2 and 3  2 are

eigenvalues of A .
1
 
For   1 the eigenspace is spanned by v1   1
8
 

0
 
For   2 the eigenspace is spanned by v2   0 
1
 
Thus A is not diagonalizable since A is 3  3 matrix and there are only two linearly
independent vectors.

Theorem 1.2.4 If an n  n matrix A has n distinct eigenvalues then A is diagonalizable.

Caution: Just because the n  n matrix A has fewer than n distinct eigenvalues,
you may not conclude that it is not diagonalizable. See example 1.2.2 (c).
Higher Power of a Diagonalizable Matrix
Powers of a diagonal matrix are easy to compute, and when a matrix is diagonalizable, it is
almost as easy.

Theorem 1.2.5 If A is diagonalizable matrix and P1 AP  D , where D is a diagonal matrix.

Then An  PDn P1 for any positive integer n .

Proof. The proof is by induction.

For n  1, A  PDP1 by hypothesis.

Assume for k  1 , that is, Ak  PDk P1 need to show for k  1 , that is, Ak 1  PDk 1 P1

Ak 1  Ak A   PDk P 1  PDP1  induction assumption

 PDk  P 1 P  DP 1  PDk DP 1  PD k 1 P 1

22 | BDU,DEPARTEMENT OF MATHEMATICS
 CHAPTER-1 EIGENVALUES AND EIGENVECTORS

 1 1 1
 
Example 1.2.4 Given A   1 3 1  calculate A20 .
 3 1 1
 

 1 1 1  2 0 0 
   
Solution. See example 1.2.2 (b) we have P   1 0 1  and D   0 2 0 
 4 1 1   0 0 3
   

1 1
1
 1 1 1  2
20
0 0  5 5
 1   
Thus A  PD P   1 0 1   0
20 20
220 0   0 1 1
 4 1 1   0 320   1 1
  0
 1 
5 5

 697147165 3486784401 698195741

 
  697147165 3487832977 697147165 
 696518019 3489930131 697566595 
 
This is much simpler than calculating A20 directly.

EXERSCISES 1.4_________________________________________________
1. For each of the following matrix A , find all eigenvalues and describe the eigenspace of
the matrix A . Find if possible an invertible matrix P and a diagonal matrix D such that
A  PDP1 . Otherwise, explain why A is not diagonalizable.
1 1 0  2 1 1
1 2    
a) A    b) A   0 1 1  c) A   0 3 2 
 4 3 0 0 1  1 1 2 
   
0 0 0 1  2 1 2 4 
   
0 0 1 0 12 1 4 9 
b) A   e) A  
0 1 0 0  6 5 2 4 
   
1 0 0 0  3 4 5 10 
2. Compute An , where n is a positive integer.
 1 7 1   3 1 0 
   
a) A   0 1 0  b) A   1 2 1
 0 15 2   0 1 3 
   

BDU,DEPARTEMENT OF MATHEMATICS | 23
 Math 2042 LINEAR ALGEBRA II

3. Prove theorem 1.2.4

1 a 1
 
4. Consider A   0 1 b 
0 0 c
 
a) Find the characteristic polynomial of A .
b) Explain whether A is diagonalizable when a  0, c  1 .
c) Explain whether A is diagonalizable when a  0, c  1.
d) Explain whether A is diagonalizable when c  1 .
5. In exercise a-d determine whether the statements are true or false.
a) Every n  n matrix is diagonalizable.
b) An n  n matrix A is diagonalizable if and only if there is a basis of  n
consisting of eigenvectors of A .
c) The standard vectors are eigenvectors of a diagonal matrix.
d) Every diagonalizable n  n matrix has n distinct eigenvalues.
6. If A is diagonalizable matrix, show that At is diagonalizable.
7. If A is invertible and diagonalizable matrix then show that A1 is diagonalizable.
8. Suppose A and B are two n  n matrices and there exists a single matrix P such that
P 1 AP and P 1 BP are both diagonal matrices. Show that AB  BA .
9. Suppose that A and B are similar matrices. Show that A3 and B 3 are similar matrices.
Generalize the result.
10. Suppose that A and B are similar matrices, with A non singular. Prove that B is
non singular, and that A1 is similar to B 1 .
___________________________________________________________________

1.3 Polynomials and Matrices

Let P( x)  an x n  an1x n1  ...  a1x  a0 be a polynomial function and A   aij  be square
nn

matrix. P( A)  an An  an  An1  ...  a1 A  a0 I n is a polynomial of matrix A .

If P( A)  0 then A is the root of the polynomial matrix, we say P annihilates A .

24 | BDU,DEPARTEMENT OF MATHEMATICS
 CHAPTER-1 EIGENVALUES AND EIGENVECTORS

 1 1
Example 1.3.1 Consider P( x)  3x 2  2 x  5 and A   
2 0 
 10 1 
Then P( A)  3 A2  2 A  5I 2   
 2 11
 1 0   1 0 
Let A    . Verify that P( A)  0 . Thus A    is the root of the
 0 1  0 1
polynomial P( x) .

Let f and g be polynomials over M nxn matrix. Let A be a square matrix then

a) ( f  g )( A)  f ( A)  g ( A)
b) ( fg )( A)  f ( A) g ( A)  g ( A) f ( A)
c) (kf )( A)  kf ( A) for any scalar k .

Theorem 1.3.1 Suppose A is a square matrix and  is an eigenvalue of A . Let P( x) be a

polynomial, P( x)  a0  a1 x    am x m . Then P( ) is an eigenvalue of the matrix P( A) .

Proof. Let v  0 be an eigenvector of A corresponding to eigenvalue  and

P( A)  a0 I n  a1 A    am Am .

Then P( A)v  (a0 I n  a1 A    am Am )v.

 a0 I n v  a1 Av  ...  am Am v

 a0 ( I n v)  a1 ( Av)  ...  am ( Am v)
 a0 v  a1 ( v)  ...  am ( m v)
  a0  0  a1 x  ...  am  m  v
 p ( )v

Theorem 1.3.2 Let A be a square matrix of size n  n . Then there exists a non zero
polynomial P( x) in which P( A)  0 .

Proof. All n  n matrix from a vector space of dimension n 2 under the usual addition of

matrix and scalar multiplication. Hence the n2  1 matrices, I , A, A2 ,, An is linearly

2

dependant. There fore, there exists scalars a0 , a1 ,, an not all zero such that

BDU,DEPARTEMENT OF MATHEMATICS | 25
 Math 2042 LINEAR ALGEBRA II

a0 I  a1 A  a2 A2  ...  an2 x n  0 . Thus define P  x   a0 x  a1 x  ...  an2 x n

2 2

Then it follows that P( A)  0 .

Among all the nonzero polynomials in which P( A)  0 consider those of least degree and
from them select one whose leading coefficient is 1, called monic polynomial.

Theorem 1.3.3 There is only one monic polynomial of least degree in which P( A)  0

Proof. Suppose P1 ( x) and P2 ( x) are two such polynomials.

Define g  x   P1  x   P2  x  .

If g ( x)  0 then degree of g ( x) is less than degree of P1 ( x) .

g ( A)  P1 ( A)  P2 ( A)  0 contradiction to the choice of P1 ( x) and P2 ( x) .

Definition 1.3.1 Let A be n  n matrix over a field F . The unique monic polynomial of
least degree in which P( A)  0 is called minimal polynomial of A and denoted by mA ( x)
.

Theorem 1.3.4 Let A be an n  n matrix over a field F . A is non singular if and only if
the constant term of minimal polynomial of A is non-zero.

Proof. Let mA ( x)  xt  at 1 xt 1  ...  a1 x  a0 be the minimal polynomial of A .

 Suppose A is non singular 0  mA ( A)  At  at 1 xt 1  ...  a1 A  a0 I n

If a0  0 then 0  A( At 1  at 1 At 2    a1 I n ) divide both sides by A1

Hence At 1  at 1 At 2  ...  a1 I n  0 contradiction to the minimality of mA ( x)

Hence a0  0.

  Suppose a0  0
0  mA  A  At  at 1 At 1  ...  a1 A  a0 I n
1 t 1
In 
a0
 A  at 1 At 2    a2 A  a1 I n  A
1 t 1
A1 
a0
 A  at 1 At 2    a2 A  a1 I n  . Hence A is non-singular.

This is another way to find A1 .

26 | BDU,DEPARTEMENT OF MATHEMATICS
 CHAPTER-1 EIGENVALUES AND EIGENVECTORS

Theorem 1.3.5 The minimal polynomial mA  x  of a matrix A divides every polynomial

in which P( A)  0 .

Proof. Suppose P( x) is a polynomial in which P( A)  0 . By division a logarithm there

exists polynomials q( x) and r ( x) such that

P( x)  mA ( x)q( x)  r ( x) , with degree r ( x) less than degree mA ( x) or r ( x)  0

P  A  mA  A q  A  r  A , since P( A)  0 and mA ( A)  0

We obtain r  A  0. If r  x   0 contradiction to minimality of mA ( x)

Thus r ( x)  0.

Theorem 1.3.6 (Cayley-Hamilton theorem)

Every square matrix A is a root of its own characteristic polynomial.
More specifically, if A is an n  n matrix and P     det   I n  A is the characteristic

polynomial of A , then P( A)  0 .

a b 
Example 1.3.2 Verify Cayley-Hamilton theorem for 2  2 matrix A   
c d 
Solution. PA  x   x2   a  d  x   ad  cd 

 a 2  bc ab  bd  a b  0 0
PA  A      a  d      ad  cd  I   
 ac  cd bc  d 
2 2
c d  0 0

Proof. det   I n  A   n  b1 n1  b2 xn2    bn1  bn is a polynomial of degree n .

 
Let B( )  bij ( ) be the adjoint of   I n  A . Since the bij ( ) are cofactors of the

matrix   I n  A , they are polynomials of degree at most n  1 .

Thus bij     bijn 1  n1  bijn  2  n2    bij1   bij0

Let Bk  bijk  , for k  0,1,, n  1 . Then we have

B     Bn1 n1  Bn2  n2    B1  B0

But   I n  A adj   I n  A  adj   I n  A  I n  A  det   I n  A I n

We have   I n  A B( )  det   I n  A I n . Thus

BDU,DEPARTEMENT OF MATHEMATICS | 27
 Math 2042 LINEAR ALGEBRA II

  I n  A (Bn1 n1  Bn2  n2    B1  B0 )    n  b1 n1  b2 xn2    bn1  bn  I n

Expanding the left-hand side of this equation and equating like powers of  , we have
I n  Bn 1
b1 I n  Bn  2  ABn 1
b2 I n  Bn 3  ABn  2
 
bn 1 I n  B0  AB1
bn I n   AB0

multiplying the k th row by Ar k 1 yields

An  An Bn 1
b1 An 1  An 1 Bn  2  An Bn 1
b2 An  2  An  2 Bn 3  An 1 Bn  2
 
bn 1 A  AB0  A2 B1
bn I n   AB0

Add down words An  b1 An1  b2 An2    bn1 A  bn I n  0 , that is, P( A)  0 .

Example 1.3.3 If  is the root of the characteristic polynomial of A then  is the root of the
minimal polynomial of A .
Proof. If  is the root of the characteristic polynomial of A then  is an eigenvalue of A .
Thus there is a non zero vector v  F n such that Av   v
Let mA ( x)  am xm  am1 x m1    a1 x  a0 be the minimal polynomial A .
m m
0  mA ( A)v   am A v   am  m v  mA ( )v
m

i 0 i 0

Since v  0 we have mA ( )  0 .

C Corollary 1.3.7 Let A be a square matrix, then the minimal polynomial of A divides
the characteristic polynomial of A .

28 | BDU,DEPARTEMENT OF MATHEMATICS
 CHAPTER-1 EIGENVALUES AND EIGENVECTORS

C Theorem 1.3.8 The minimal polynomial of A and the characteristic polynomial of A

have the same irreducible factors.

More specifically, if P  x    x  1  1  x  2  2 ...  x  k  k is the characteristic polynomial of

n n n

A , where 1 , 2 ,, k are distinct eigenvalues. Then there exists integers m1 , m2 ,, mk such that

mA  x    x  1   x  2   x  k 
m1 m2 mk
with 1  mi  ni , for i  1, 2,, k .

Proof. Let A  (aij )nn , since the minimal polynomial divides the characteristic polynomial it

suffices to show xI n  A divides mA ( x)n .

Let mA ( x)  xr  c1 xr 1  c2 x r 2    cn1 x  cn be the minimal polynomial of A

Consider the following matrices
B0  I n
B1  A  c1 I n
B2  A2  c1 A  c2 I n
(1)
B3  A3  c1 A2  c2 A  c3 I n
 
r 1
Br 1  A  c1 Ar  2  c2 Ar 3    cr 1 I n
Then
B0  I n
B1  AB0  c1 I n
B2  AB1  c2 I n
(2)
B3  AB2  c3 I n
 
Br 1  ABr  2  cr 1 I n

multiplying the k th row by Ar k 1 yields

Ar B0  Ar
Ar 1 B1  Ar B0  c1 Ar 1
Ar  2 B2  Ar 1 B1  c2 Ar  2
(3)
Ar 3 B3  Ar  2 B2  c3 Ar 3
 
ABr 1  A Br  2  cr 1 A
2

BDU,DEPARTEMENT OF MATHEMATICS | 29
 Math 2042 LINEAR ALGEBRA II

Add downwards
ABr 1  Ar  c1 Ar 1  c2 Ar 2    cn1 A (4)

Add and subtract cr I n in equation (4)

ABr 1  ( Ar  c1 Ar 1  c2 Ar 2    cn1 A  cr I n )  cr I n

ABr 1  cr I n since Ar  c1 Ar 1  c2 Ar 2    cn1 A  cr I n  mA ( x)  0

Let B( x)   xr 1 B0  x r 2 B1    xBr 2  Br 1 then

 xI n  A B( x)  ( x r B0  x r 1 B1    xBr 1 )  ( x r 1 AB0  x r 2 AB1    xABr  2  ABr 1 )

 x r B0  x r 1 ( B1  AB0 )  x r 2 ( B2  AB1 )    x( Br 1  ABr  2 )  ABr 1
 x r I n  x r 1c1 I n  x r  2 c2 I n    xcr 1 I n  cr I n
 mA ( x ) I n

Thus  xI n  A B( x)  mA ( x) I n .

Hence det  xI n  A det  B( x)   det  mA ( x) I n   mA ( x)n .

Since det( B( x)) is a polynomial it follows that xI n  A divides mA ( x)n .

C Corollary 1.3.9 If the eigenvalues of A are the distinct then the minimal polynomial of A
and the the characteristic polynomial of A are identical.

 1 0 0
 
Example 1.3.4 Given A   1 2 0  then find
 3 5 2 
 
a) the minimal polynomial of A . b) the inverse of A . c) A3 .

Solution. a) The characteristic polynomial P     det   I3  A     1   2  . Thus by

2

theorem 1.3.7 the first candidate for the minimal polynomial is mA        1   2  .

 0 0 0  1 0 0   0 0 0 
    
mA  A   I  A 2 I  A    1 1 0  1 0 0    0 0 0   0
 3 5 1 3 5 0   5 5 0 
    
mA        1   2 

30 | BDU,DEPARTEMENT OF MATHEMATICS
 CHAPTER-1 EIGENVALUES AND EIGENVECTORS

Hence mA  x   PA  x   1    2   
2

b) mA  x   PA  x   1    2    =  3  5 2  8  4 . By theorem 1.3.4 A has an

2

inverse. mA  A  A3  5 A2  8 A  4I3  0 . It follows I 3 

4
 A  16 A  8I3  A
1 2

 4 0 0
Hence A1   A2  5 A  8I 3    2 2 0 
1
4  11 5 2 
 

1 0 0
c) From (b) we have A  5 A  8 A  4 I 3   7 8 0 
3 2

 4 60 8 
 

 1 1 1
 
Example 1.3.5 Find the minimal polynomial of A   1 3 1 
 3 1 1
 
Solution. The characteristic polynomial
P     det   I3  A   3  3 2  4  12

    2   2  (  3)  mA ( ) by corollary 1.3.9

C Theorem 1.3.10 Let A be a square matrix of size n  n . A is diagonalizable if and only if

the minimal polynomial of A is the form mA  x    x  1  x  2  x  k  , where the

i ’s are the distinct eigenvalues of A .

Proof.    Suppose A is diagonalizable and 1 , 2 ,, k are the distinct eigenvalues of A .

Define g  x    x  1  x  2  x  k  then g  x  divides mA  x  the minimal polynomial of

A by theorem 1.3.8. Let   v1 , v2 ,, vn  be a basis of F n then for any j  1, 2,, n we have

 i I n  A v j  0 for some eigenvalue i , i  1, 2,, k .

Since  x  i  divides g  x  , for any i  1, 2,, k . It follows g  x   qi ( x)( x  i )

g  A v j  qi  A A  i I n  v j  0 , for any j  1, 2,, n

Hence g  A  0 , since g  A takes each element of a basis in to 0.

BDU,DEPARTEMENT OF MATHEMATICS | 31
 Math 2042 LINEAR ALGEBRA II

Therefore g  x  is the minimal polynomial of A .

(  ) Suppose mA  x    x  1  x  2  ...  x  k  with  i ' s are eigenvalues of A .

Write mA  x    x   j  Pj  x  where Pj  x  is a polynomial of degree less than k .

Claim      Im P ( A) , where Im P ( A) is the image set of P ( A)

A j j j j

To show Im P ( A)     
j A j

Let w  Im Pj ( A) then w  Im Pj ( A)v , for some v  F n .

0  mA  A v   A   j I n  Pj  A  v
  A   j In  w

In other words Aw   j w .

Thus w   A   j  , for each j  1, 2,, k

To show      Im P ( A)
A j j

Let v      , so we have Av   v
A j j

It follows that Pj ( A)v  Pj ( j )v

But Pj ( j )  0 otherwise  x   j  divides mA  x 

2


v  Pj ( A)  Pj ( j )  v
1

Hence v  Im Pj ( A)


k
Finally claim j 1
ImPj ( A)  F n

To see this the polynomial P( x)   j 1  Pj ( j )  Pj ( x) has degree less than k .

k 1

P( x) assumes the value 1 at each k distinct elements 1 , 2 ,, k of F .

There is only one such polynomial and this is P( x)  1 .

For each v  F n , we have

v  I n v  P( A)v   j 1  Pj ( j )  Pj ( A)v
k 1

Thus v   j 1 Im Pj ( A)
k

32 | BDU,DEPARTEMENT OF MATHEMATICS
 CHAPTER-1 EIGENVALUES AND EIGENVECTORS

Both claim together implies that  j 1 dim A   j   n

k

Example 1.3.6 Let A be a 2  2 matrix such that A2  3 A  2I  0 . Then show that A is

diagonalizable.
Solution. Let mA  x  be the minimal polynomial of A and P( x)  x 2  3x  2

Then mA  x  divides ( x  1)  x  2  The possible candidates of mA  x  are

mA  x   x  1, mA  x   x  2 or mA  x    x  1 x  2 . In any case A is diagonalizable

More specifically If mA  x   x  1 then A  I 2 .

If mA  x   x  2 then A  2 I 2 .

1 0
If mA  x    x  1 x  2  then A is similar to  .
0 2

EXERSCISES 1.5_________________________________________________
 4 3
1. Let P( x)  x 2  3x  2 and A   .
 2 1
a) Evaluate PA     det( I 2  A) b) Verify that PA  A  0
2. Find the minimal polynomial of
a) the identity matrix? b) the zero matrix?
0 1 0
 
3. Let A   0 0 1  then find
 1 3 3 
 
a) the characteristic polynomial of A b) the minimal polynomial of A
c) the inverse of A d) A3
0 1 0 1
 
 1 0 1 0
4. Let A  then
0 1 0 1
 
1 0 1 0
a) find the characteristic polynomial of A b) find the minimal polynomial of A
c) show A  4 A
3
d) find the inverse of A
5. Prove a) corollary 1.3.7 b) corollary 1.3.9
6. Show that if det( A)  0 show that A1 can be obtained as a polynomial in A.

BDU,DEPARTEMENT OF MATHEMATICS | 33
Math 2042 LINEAR ALGEBRA II

7. Find the minimal polynomial of the following matrices.

2 1 0 0  1 1 0 0
0 0 c     1 1 0 0 
a) A  1 0 b 
0 2 0 0
b) A   c) A   
0 0 1 1  2 2 2 1 
0 1 a     
 0 0 2 4   1 1 1 0 
8. Show that a nilpotent matrix is diagonalizable if and only if it is the zero matrix.

1 1 0 2 0 0
   
9. i) Let A   0 2 0  and B   0 2 2  show that
0 0 1  0 0 1
  
a) A and B are not similar b) have the same minimal polynomial

2 1 0 0 2 0 0 0
   
0 2 0 0 0 2 1 0
ii) Let A   and B   the find
0 0 2 1 0 0 2 1
   
0 0 0 2 0 0 0 2
a) the characteristic polynomials of A and B
b) the minimal polynomials of A and B

x 1 0  0 0
 
0 x 1  0 0
10. Let A       
 
0 0 0  x 1
0 0 0  0 x 

Show that the characteristic polynomials is PA        x 
n

11. Show that a square matrix A and its transpose have the same minimal polynomial.
12. Let A be an n  n matrix such that A3  A . Show that A is diagonalizable.
13. The most common incorrect proof of Cayley-Hamilton theorem is the following.
We know characteristic polynomial of A , PA ( x)  det( I n  A)
Thus PA ( A)  det( AI n  A)  det  0   0 .
What is wrong in the above argument?
14. If D is a diagonal matrix whose diagonal elements are 1 , 2 ,, n and f ( x) is a
polynomial, then f ( D) is a diagonal matrix whose diagonal elements are
f (1 ), f (2 ),, f (n ) correspondingly.

34 | BDU,DEPARTEMENT OF MATHEMATICS
 CHAPTER-1 EIGENVALUES AND EIGENVECTORS

15. Evaluate e A , where

1 0 0
 1 1  
a) A    b) A   2 2 0 
 1 2   2 1 3 
 
16. Let P( x) be a polynomial.
a) If  is an eigenvalue of A and P( A)  0 then show that P( )  0 .
b) If 1 , 2 ,, n are eigenvalues of A then show that P(1 ), P(2 ),, P(n )
are eigenvalues of P( A)
___________________________________________________________________

1.4 Eigenvalues and Eigenvectors of Linear Operators

So far we have discussed eigenvalues and eigenvectors for an n  n matrix A .
Then the eigenvalues and the eigenvectors of a linear operator T can also be defined as follows.

C Definition 1.4.1 Let T : V  V be a linear operator on a vector space V . The eigenvectors of

of T are the non-zero vector solutions of equation
T (v)   v (1)
where  is a scalar called an eigenvalue of the linear operator T .

Example 1.4.1 Let I : V  V be an identity operator on a vector space V .

For each v  V  0 , I (v)  v  1.v .

Thus   1 is an eigenvalue of the identity operator for every non-zero eigenvector v  V .

Example 1.4.2 Let T : P1  P1 be an operator defined by T  a  bx    a  b   (a  b) x .

  0 is an eigenvalue of T corresponding to an eigenvector of the form a  ax, a  0 .

  2 is an eigenvalue of T corresponding to an eigenvector of the form a  ax, a  0

Example 1.4.3 Consider the vector space V  C  ( ) .

a) the vector f ( x)  eax , with a  0 , is an eigenvector with eigenvalue a of the linear

df
operator T : V  V , given by T ( f ) 
dx

BDU,DEPARTEMENT OF MATHEMATICS | 35
 Math 2042 LINEAR ALGEBRA II

b) the vector f ( x)  cos(ax) , with a  0 , is an eigenvector with eigenvalue a 2 of the linear

d2 f
operator T : V  V , given by T ( f ) 
dx 2
How to Solve Eigenvalue Problem for Operators?
1. Recall that If T : V  V is a linear operator on n -dimensional vector space V then there
exists an n  n matrix A such that T (v)  Av .
2. On the other hand the matrix representation of a linear operator T : V  V depend on
the basis vectors selected for V . Let T : V  V be a linear operator on n -dimensional
vector space V . If   v1 , v2 ,, vn  and  '  v1 , v2 ,, vn  are two bases for V and A

and A ' are the matrix representation of T relative to  and  ' respectively, then A
and A ' are similar. Hence eigenvalues of A and A ' are equal.
(See exerscise 1.6 problem 4)
3. Select your favorite basis,   v1 , v2 ,, vn  and find the matrix representation of T relative

to  , usually denoted by T  . Then T (v)  u if and only if Av   u  , where  v  is

the coordinates or coefficients of v in basis  .

It follows that
T (v)   v, v  0  Av   v  , v 0 (2)

   I n  A v  0, v 0

Hence the eigenvalue problems(1) and (2) are equal with  v  the coordinates of v in basis  .

Thus the eigenvalue and eigenvectors of a linear operator T :  n   n are just the
eigenvalue and eigenvectors of its matrix representation in standard basis.
In general as we see from (2) the operator T : V  V and its matrix representation A in any
basis  of V have the same eigenvalue  but the eigenvectors of A are the coordinates of the
eigenvectors of T in basis  .
Hence every theorem about a matrix corresponds to a theorem about a linear operator.
Remark: If T : V  V is a linear on a vector space V the powers of T defined by the
composition operation, that is,
T 2  T 0T , T 3  T 2 0T , 

36 | BDU,DEPARTEMENT OF MATHEMATICS
 CHAPTER-1 EIGENVALUES AND EIGENVECTORS

Also for any polynomial P( x)  an x n  an1 x n1   a1 x  a0 , we define P(T ) in the same

way as we did for matrices, that is, P(T )  anT n  an1T n1   a1T  a0 I n
Example 1.4.4 Find the eigenvalues and a basis for the eigenspace of the linear operator
T :  2   2 , given by T  x, y    2 x  5 y,6 x  y  .

Solution. The matrix representation of T in a standard basis   (1, 0),  0,1 is

 2 5
T   T (1, 0) T (0,1)    A
 6 1
The characteristic equation det( I 2  A)  0 for A and therefore for T is

(  7)    4   0

The roots of this equation, hence the eigenvalues of A and T are 1  7, and 2  4
Corresponding eigenvector of A and T
For 1  7 ,  v1    1,1

For 2  2 , v2    5, 6  . Verify this.

Definition 1.4.2 A linear operator T on a finite dimensional vector space V is called

diagonalizable if there exist a basis  of V such that T  is a diagonalizable matrix.

Example 1.4.5 Find the eigenvalues and a basis for the eigenspace of the linear operator

 
T : P2  P2 , defined by T a  bx  cx 2  2c   a  2b  c  x  (a  3c) x 2 .

Solution. The matrix representation of T in a standard basis   1, x, x 2  is

 0 0 2 
T  
 T (1)  T ( x) T ( x ) 
2
   
 1 2 1   A
1 0 3 
 
The characteristic equation det( I3  A)  0 for A and therefore for T is

(  2)   2  3  2   0

The roots of this equation, hence the eigenvalues of A and T are 1  1, and 2,3  2

BDU,DEPARTEMENT OF MATHEMATICS | 37
 Math 2042 LINEAR ALGEBRA II

Corresponding eigenvector of A
For 1  1 , v1    2,1,1

For 2,3  2 , v2    1, 0,1 and v3    0,1, 0  Verify.

The vectors  v1  ,  v2  and  v3  are coordinates of v1 , v2 and v3 in basis   1, x, x 2 

Thus the basis of eigenspace of T corresponding to

1  1 , with v1  2  x  x 2

2,3  2 , with v2  1  x2 and v3  x .

T has three linearly independent eigenvectors. Thus T is diagonalizable.

If T : V  V is a linear operator on a finite dimensional vector space, we have seen that T has a
matrix representation. The matrix representation of T depend on the basis vectors selected for V .
Thus our next question is that, how to choose a basis for V that make matrix representation of
T as simple as possible?

Theorem 1.4.1 A linear operator T on a finite dimensional vector space V is diagonalizable if

and only if there is a basis   v1 , v2 ,, vn  for V and scalars 1 , 2 ,, n (not necessarily

distinct) such that T (v j )   j v j for 1  j  n . In this case

 1 0 0  0
 
0 2 0  0
T  
    
 
0 0 0  n 

Proof.
   Suppose T is diagonalizable there exists a basis  of V such that T   D is a diagonal

matrix. Let D  (dij )nn with d jj   j and   v1 , v2 ,, vn  then for each j , 1  j  n

T (v j )  i 1 di j vi  d jj v j   j v j
n

38 | BDU,DEPARTEMENT OF MATHEMATICS
 CHAPTER-1 EIGENVALUES AND EIGENVECTORS

  If there is a basis   v1 , v2 ,, vn  for V and scalars 1 , 2 ,, n such that

T (v j )   j v j for 1  j  n .

 1 0 0  0
 
2
 T (vn ) 
0 0  0
T   T (v1 ) T (v2 ) T (v3 )  
    
 
0 0 0  n 

Example 1.4.6 Consider the linear operator T : P2  P2 defined by

T  p( x)   p( x)  ( x  1) p '( x) . Determine whether T is diagonalizable or not. If T is

diagonalizable, find the basis that make T diagonalizable.

Solution. The matrix representation of T in a standard basis   1, x, x 2  is

1 1 0
 
T   0 2 2  A
0 0 3
 
The characteristic equation det( I3  A)  0 for T is (  1)(  2)    3  0

The eigenvalues of T are   1,   2 and   3

Hence T is diagonalizable. (Verify)
For 1  1 , corresponding eigenvector of T is v1  p( x)  1.

For 2  2 , corresponding eigenvector of T is v2  p( x)  1  x

For 3  3 , corresponding eigenvector of T is v3  p( x)  1  2 x  x2

The set of eigenvector of T ,  '  1,1  x,1  2 x  x 2  a basis of P2 . Verify.

1 0 0
 
T  '   0 2 0  which is a diagonal matrix.
 0 0 3
 

Example 1.4.7 Consider the linear operator T :  3   3 give by

T  x, y, z    2 z, x  2 y  z, x  3z  . Determine whether T is diagonalizable or not. If T is

diagonalizable, find the basis that make T diagonalizable.

Solution. The matrix representation of T in a standard basis is

BDU,DEPARTEMENT OF MATHEMATICS | 39
 Math 2042 LINEAR ALGEBRA II

 0 0 2 
T    1 2 1   A
1 0 3 
 

The characteristic equation det( I3  A)  0 for T is (  1)(  2)2  0

The eigenvalues of T are   1 and   2

Corresponding eigenvector are
for 1  1 , v1  (2,1,1) .

for 2  2 , v2  (0,1, 0) and v3  (1,0,1) . There are three linearly independent eigenvectors of

T . Hence T is diagonalizable.
The set of eigenvector of T ,    2,1,1 , (0,1, 0),  1, 0,1 a basis of  3 .

1 0 0
T  
 T (v1 ) T (v2 ) T (v3 )   
  0 2 0  a diagonal matrix.
0 0 2
 
Verify that T (v j )   j v j , for j , 1  j  3 .

EXERSCISES 1.6_________________________________________________
1. For the following linear operators.
i ) Write the matrix representation of A  T in a standard basis.
ii) Write the characteristic equation of T (of A ) in a standard basis.
iii) Find the eigenvalues of T .
iv) Find the basis for each of the eigenspace of T .
a) T :  3   3 defined by T ( x, y, z)  ( x  z,3 y, z  x) .

b) T : P2  P2 defined by T (a  bx  cx2 )   4a  c    b  2a  x   c  2a  x 2

  a b   10a  9b 4a  2b 
c) T : M 22  M 22 defined by T     
  c d    2c  5d c  2d 

2. If T linear transformation which rotates all vectors in  2 through

300, explain why T can’t have any real eigenvalue.
3. Find all linear operators in 3 -dimensional vector space which is diagonalizable and
5 is the only eigenvalue.

40 | BDU,DEPARTEMENT OF MATHEMATICS
 CHAPTER-1 EIGENVALUES AND EIGENVECTORS

4. Let T : V  V be a linear operator on n -dimensional vector space V .

If   v1 , v2 ,, vn  and  '  v1 , v2 ,, vn  are two bases for V then show that

T  '  P1 T  P , where P transition matrix from  ' to  .

More specifically, the two matrix T  and T  ' are similar.

(Hint: Consider the composition operator T  '  I  '  0 T  0I  ' )

I T I

Basis  ' Basis  Basis  Basis  '

5. Let T :  2   2 be a linear operator defined by T ( x, y)  ( x  y, x  y) and

P(t )  t 3  t 2  2t  5 . Find P(T ( x, y)) .
6. Find the minimal polynomial for each of the following linear operator
a) T :  3   3 defined by T ( x, y, z)  (5x  3 y  2 z,6 x  4 y  4 z, 4 x  4 y  5z)
b) T : P2  P2 defined by T ( p)  p ' 2 p .
c) T : M nn  M nn defined by T ( A)  At .

7. For the following linear operators T . Determine whether T is diagonalizable or not.

If T is diagonalizable, then find the basis  such that T  is a diagonalizable matrix.
a) T :  2   2 defined by T ( x, y)  ( x  y, x  3 y)
b) T :  3   3 defined by T ( x, y, z)  (5x  3 y  2 z,6 x  4 y  4 z, 4 x  4 y  5z)
c) T : P2  P2 defined by T ( p)  p(0)  p(1)( x 2  x) .
d) T : P2  P2 defined by T ( p)  p '( x) .
e) T :  3   3 defined by T ( x, y, z)  (3x  y,  x  2 y  z,  y  3z)
f) T : M 22  M 22 defined by T ( A)  At .
8. Let T : V  V be a linear operator and   v1 , v2 ,, vn  be a basis of V such that
T (v j )   j v j for 1  j  n . Suppose v   i 1 xi vi then show that
n

T (v)  i 1 xi i vi
n
a)

b) T n (v)  i 1 xi i n vi , for each positive integer n .

n

9. Let T :  3   3 have three distinct positive eigenvalues. Show that there is a linear
operator S :  3   3 such that S 2  T .
_______________________________________________________

BDU,DEPARTEMENT OF MATHEMATICS | 41
 Math 2042 LINEAR ALGEBRA II

Summery Exercise
1. Let A be n  n matrix. True or false:
a) At has the same eigenvalues as A .
b) At has the same eigenvectors as A .
c) If A is diagonalizable, then so is At .
2. Consider the matrix
 1 1 6 
 
A  0 2 3 
 0 0 1 
 
a) Find its eigenvalues by inspection.
b) Is matrix A diagonalizable?
c) If the matrix is diagonalizable, diagonalize it.
3. Show that if AX   X and AY  Y then whenever  and  are scalars,
A( X   Y )   ( X   Y )
Does this imply that  X   Y is an eigenvector? Explain.
4. Show that the following are equivalent for n  n matrix A and   
a)  is an eigenvalue of A
b) there exists v  0 such that Av   v
c) there exists v  0 such that ( I n  A)v  0
d)  is a solution to det( I n _ A)  0
e) ( I n  A)v  0 has infinitely many solutions.
5. Are the matrices
1 0 0  5 0 0 
   
 0 5 0  and  0 1 0  similar?
 0 0 2   0 0 2 
   
6. If A and B are diagonalizable and have the same eigenvalues show that
A and B are similar matrix.
7. Definition: Given a vector space V and a linear operator T : V  V , the subspace
W in V is invariant under T if and only if for each w  W , T (w) W , that is,
T (W )  W . If  is an eigenvalue of T show that the eigenspace of T for  ,  T ( ) , is
an invariant subspace of the vector space V under the operator T .
8. Let A be n  n matrix
a) For every eigenvalue A of the geometric multiplicity cannot exceed its algebraic
multiplicity.
b) A diagonalizable if and only if for every eigenvalue A the geometric multiplicity
is equal to the algebraic multiplicity.

42 | BDU,DEPARTEMENT OF MATHEMATICS