EIGENVALUES AND EIGENVECTORS

ELECTRONIC VERSION OF LECTURE

Dr. Lê Xuân Đại

HoChiMinh City University of Technology
Faculty of Applied Science, Department of Applied Mathematics
Email: ytkadai@hcmut.edu.vn

HCMC — 2018.
Dr. Lê Xuân Đại (HCMUT-OISP) EIGENVALUES AND EIGENVECTORS HCMC — 2018. 1 / 56
OUTLINE

1 THE REAL WORLD PROBLEMS

2 EIGENVALUES AND EIGENVECTORS OF A MATRIX

3 DIAGONALIZATION

4 MATL AB

Dr. Lê Xuân Đại (HCMUT-OISP) EIGENVALUES AND EIGENVECTORS HCMC — 2018. 2 / 56

 The Real World Problems

MODELLING MOTION

4PQR → 4P 0 Q0 R0 is
the reflection over the x−axis.
Dr. Lê Xuân Đại (HCMUT-OISP) EIGENVALUES AND EIGENVECTORS HCMC — 2018. 3 / 56
 The Real World Problems
à !
1 0
A= is the reflection matrix.
0 −1
Therefore, for every point in the plane
(x1 , x2 ), the matrix that results in a reflection
over the x−axis and then we obtain a new
point
à ! inà the plane ! à (y1!, y2 )à !
y1 1 0 x1 x1
= . =
y2 0 −1 x2−x2
Question:
à ! ÃFor !every point (x1 , x2 ), find
y1 x1
= Ak . , (k ∈ N).
y2 x2
Dr. Lê Xuân Đại (HCMUT-OISP) EIGENVALUES AND EIGENVECTORS HCMC — 2018. 4 / 56
 Eigenvalues and Eigenvectors of a Matrix Definition
à ! à ! à !
1 0 −1 0
A= , u= ,v= . We have
0 −1 −1 1
à ! à !
−1 −1
A = and
−1 1
à ! à ! à !
0 0 0
A = = −1.
1 −1 1

Dr. Lê Xuân Đại (HCMUT-OISP) EIGENVALUES AND EIGENVECTORS HCMC — 2018. 5 / 56

 Eigenvalues and Eigenvectors of a Matrix Definition

DEFINITION 2.1
If A is an n × n matrix, then a nonzero vector
X ∈ Rn , X 6= 0 is called an eigenvector of A if
AX = λ.X for some scalar λ. The scalar λ is
called an eigenvalue of A and X is said to be
an eigenvector corresponding to λ.

EXAMPLE 2.1
Find
à eigenvalues
! and eigenvectors of
1 0
A=
0 −1
Dr. Lê Xuân Đại (HCMUT-OISP) EIGENVALUES AND EIGENVECTORS HCMC — 2018. 6 / 56
 Eigenvalues and Eigenvectors of a Matrix Definition

The equation AX = λX can be rewritten as

à − λI)X! Ã= 0 ! Ã
(A !
1 0 x1 λx1
= ⇔
0 −1 x2 λx2
à !à ! à !
1−λ 0 x1 0
= .
0 −1 − λ x2 0
This homogeneous linear system has
non-zero
¯
solution X 6=¯ 0, thus
¯ 1−λ 0 ¯
¯ = 0 ⇔ λ2 − 1 = 0
¯ ¯
¯ 0 −1 − λ
¯
¯
⇔ λ1 = −1, λ2 = 1.
Dr. Lê Xuân Đại (HCMUT-OISP) EIGENVALUES AND EIGENVECTORS HCMC — 2018. 7 / 56
 Eigenvalues and Eigenvectors of a Matrix Definition

In the case where λ1 = −1, we have

(
2x1 + 0x2 = 0
⇔ x1 = 0, x2 = α.
0x1 + 0x2 = 0
Therefore, the eigenvectors corresponding
to λ1 = −1 are α(0, 1), α 6= 0.
In the case where λ2 = 1. We have
(
0x1 + 0x2 = 0
⇔ x1 = β, x2 = 0.
0x1 − 2x2 = 0
Therefore, the eigenvectors corresponding
to λ2 = 1 are β(1, 0), β 6= 0.
Dr. Lê Xuân Đại (HCMUT-OISP) EIGENVALUES AND EIGENVECTORS HCMC — 2018. 8 / 56
 Eigenvalues and Eigenvectors of a Matrix Definition

EXAMPLE 2.2
Find
à eigenvalues
! and eigenvectors of
1 2
A=
−2 1

Dr. Lê Xuân Đại (HCMUT-OISP) EIGENVALUES AND EIGENVECTORS HCMC — 2018. 9 / 56

 Eigenvalues and Eigenvectors of a Matrix Definition

à !à !
1 2 x1
AX = λX can be rewritten =
−2 1 x
à ! à !à ! à ! 2
λx1 1−λ 2 x1 0
⇔ = .
λx2 −2 1 − λ x2 0
This homogeneous linear system has
non-zero solution X 6= 0, thus
¯ ¯
¯ 1−λ 2 ¯
¯ = 0 ⇔ (1 − λ)2 + 4 = 0
¯ ¯
¯ −2 1 − λ
¯
¯

⇔ λ1,2 = 1 ± 2i.
Dr. Lê Xuân Đại (HCMUT-OISP) EIGENVALUES AND EIGENVECTORS HCMC — 2018. 10 / 56
 Eigenvalues and Eigenvectors of a Matrix Definition

In the case where λ1 = 1 + 2i. We have

(
−2ix1 + 2x2 = 0
⇔ x1 = α, x2 = αi.
−2x1 − 2ix2 = 0
Therefore, the eigenvectors corresponding
to λ1 are α(1, i), α 6= 0.
In the case where λ2 = 1 − 2i. We have
(
2ix1 + 2x2 = 0
⇔ x1 = β, x2 = −βi.
−2x1 + 2ix2 = 0
Therefore, the eigenvectors corresponding
to λ2 are β(1, −i), β 6= 0.
Dr. Lê Xuân Đại (HCMUT-OISP) EIGENVALUES AND EIGENVECTORS HCMC — 2018. 11 / 56
 Eigenvalues and Eigenvectors of a Matrix Characteristic equation

If λ is an eigen value of A ⇔ ∃X 6= 0 : AX = λ.X

⇔ AX − λX = 0 ⇔ (A − λI).X = 0.
This homogeneous linear system has
non-zero solution X 6= 0, thus det(A − λI) = 0
DEFINITION 2.2
If A is an n×n matrix, then λ is an eigenvalue
of A if and only if χA(λ) = det(A − λI) = 0. This
is called the characteristic equation of A. The
polynomial χA(λ) = det(A − λI) is called the
characteristic polynomial.
Dr. Lê Xuân Đại (HCMUT-OISP) EIGENVALUES AND EIGENVECTORS HCMC — 2018. 12 / 56
 Eigenvalues and Eigenvectors of a Matrix Characteristic equation

FINDING EIGENVALUES AND EIGENVECTORS OF A

SQUARE MATRIX

STEP 1. Finding the characteristic equation

det(A − λI) = 0.
STEP 2. Solving this equation to find
eigenvalues.
STEP 3. For every eigenvalue λi , solve the
homogeneous system (A − λi I)X = 0
to find eigenvectors X
corresponding to the eigenvalue λi .
Dr. Lê Xuân Đại (HCMUT-OISP) EIGENVALUES AND EIGENVECTORS HCMC — 2018. 13 / 56
 Eigenvalues and Eigenvectors of a Matrix Characteristic equation

THEOREM

2.1 
a11 a12 a13
If A =  a21 a22 a23  , then
 
a31 a32 a33

χA (λ) = |A − λI| = −λ3 + tr(A)λ2 −

ï ¯ ¯ ¯ ¯ ¯!
¯ a a ¯ ¯a a ¯ ¯a a ¯
¯ 11 12 ¯ ¯ 22 23 ¯ ¯ 11 13 ¯
− ¯ ¯+¯ ¯+¯ ¯ λ + det(A)
¯a21 a22 ¯ ¯a32 a33 ¯ ¯a31 a33 ¯
where tr(A) = a11 + a22 + a33 is called the trace
of A.
Dr. Lê Xuân Đại (HCMUT-OISP) EIGENVALUES AND EIGENVECTORS HCMC — 2018. 14 / 56
 Eigenvalues and Eigenvectors of a Matrix Eigenspace corresponding to the eigenvalue

DEFINITION 2.3
The eigenvetors corresponding to the
eigenvalue λ, together with the zero vector,
form the null space of the matrix (A − λI).
This subspace is called the eigenspace
corresponding to the eigenvalue λ.

Dr. Lê Xuân Đại (HCMUT-OISP) EIGENVALUES AND EIGENVECTORS HCMC — 2018. 15 / 56

 Eigenvalues and Eigenvectors of a Matrix Geometric and Algebraic Multiplicity

DEFINITION 2.4
If λ0 is an eigenvalue of an n × n matrix A,
then the dimension of the eigenspace
corresponding to λ0 is called the geometric
multiplicity of λ0, and the number of times
that λ − λ0 appears as a factor in the
characteristic polynomial of A is called the
algebraic multiplicity of λ0.

Dr. Lê Xuân Đại (HCMUT-OISP) EIGENVALUES AND EIGENVECTORS HCMC — 2018. 16 / 56

 Eigenvalues and Eigenvectors of a Matrix Geometric and Algebraic Multiplicity

EXAMPLE

2.3 
3 1 1
Let A =  2 4 2 
 
1 1 3
1
Find the characteristic polynomial of A
2
Evaluate det(A − 2013.I)
3
Find eigenvalues and eigenvectors of A

Dr. Lê Xuân Đại (HCMUT-OISP) EIGENVALUES AND EIGENVECTORS HCMC — 2018. 17 / 56

 Eigenvalues and Eigenvectors of a Matrix Geometric and Algebraic Multiplicity

1. The characteristic
¯ polynomial of¯ A
¯ 3−λ 1 1 ¯¯
¯
χA (λ) = |A − λI| = ¯ 2 4 − λ 2 ¯ =
¯ ¯
1 3−λ ¯
¯ ¯
¯ 1
= −λ3 + 10λ2 − 28λ + 24 = −(λ − 2)2 (λ − 6)
2. det(A − 2013.I) = −(2013 − 2)2(2013 − 6)
3. The characteristic
¯ equation of A¯
¯ 3−λ 1 1 ¯¯
¯
χA (λ) = |A − λI| = ¯ 2 4 − λ 2 ¯ = 0
¯ ¯
1 3−λ ¯
¯ ¯
¯ 1
⇔ −(λ − 2)2 (λ − 6) = 0 ⇔ λ1 = 2, λ2 = 6.
Dr. Lê Xuân Đại (HCMUT-OISP) EIGENVALUES AND EIGENVECTORS HCMC — 2018. 18 / 56
 Eigenvalues and Eigenvectors of a Matrix Geometric and Algebraic Multiplicity

In the case where λ1 = 2, we have



 x1 + x2 + x3 = 0
2x1 + 2x2 + 2x3 = 0

 x1 + x2 + x3 = 0
   
−1 −1
⇒ X1 = α  1  + β  0  , α2 + β2 6= 0.
   
0 1
The algebraic multiplicity of λ1 = 2 is 2 and
the geometric multiplicity of λ1 = 2 is 2.

Dr. Lê Xuân Đại (HCMUT-OISP) EIGENVALUES AND EIGENVECTORS HCMC — 2018. 19 / 56

 Eigenvalues and Eigenvectors of a Matrix Geometric and Algebraic Multiplicity

In the case where λ2 = 6, we 

 have
 −3x1 + x2 + x3 = 0
 1
2x1 − 2x2 + 2x3 = 0 ⇒ X2 = γ  2  , γ 6= 0.
 

 x + x − 3x = 0
1 2 3 1
The algebraic multiplicity of λ2 = 6 is 1 and
the geometric multiplicity of λ2 = 6 is 1.

Dr. Lê Xuân Đại (HCMUT-OISP) EIGENVALUES AND EIGENVECTORS HCMC — 2018. 20 / 56

 Diagonalization Definition

DEFINITION 3.1
If A and B are square matrices, then we say
that B is similar to A if there is an invertible
matrix S such that B = S−1AS.
DEFINITION 3.2
A square matrix A is said to be
diagonalizable if it is similar to some
diagonal matrix D, that is, if there exists an
invertible matrix S such that S−1AS = D. In
this case the matrix S is said to diagonalize A.
Dr. Lê Xuân Đại (HCMUT-OISP) EIGENVALUES AND EIGENVECTORS HCMC — 2018. 21 / 56
 Diagonalization Definition

We have S−1AS = D = dig(λ1, λ2, . . . , λn). It

follows that AS = SD
λ1 0
   
a11 a12 ... a1n ... 0
a a ... a2n   0 λ ... 0 
 21 22 2
A= ,D = 
  
 ... ... ... ...  ... ... ... ...

 
an1 an2 ... ann 0 0 ... λn
 
s11 s12 ... s1n
s s ... s2n  ¡
 21 22 ¢
S=  = S∗1 S∗2 . . . S∗n

 ... ... ... ... 
sn1 sn2 ... snn
Dr. Lê Xuân Đại (HCMUT-OISP) EIGENVALUES AND EIGENVECTORS HCMC — 2018. 22 / 56
 Diagonalization Definition

   
a11 a12 ...a1n s11 s12 ... s1n
a a ...a2n   s21 s22
  ... s2n 
 21 22
AS =  .

 ... ... ......   ... ... ... ...


an1 an2 ...ann sn1 sn2 ... snn
¡ ¢ ¡ ¢
= A S∗1 S∗2 . . . S∗n = AS∗1 AS∗2 . . . AS∗n

λ1 0
  
s11 s12 ... s1n ... 0
s s22 ... s2n  0 λ ... 
 21 2
SD = 
 
 ... ... ... ...  ... ... ... ... 
 

sn1 sn2 ... snn 0 0 . . . λn

Dr. Lê Xuân Đại (HCMUT-OISP)
¡ EIGENVALUES AND EIGENVECTORS HCMC — 2018.
¢ 23 / 56
 Diagonalization Definition

Therefore,
(AS)∗i = AS∗i = (SD)∗i = λi S∗i , (i = 1, 2, . . . , n).

So, S∗i is the eigenvector corresponding to

eigenvalue λi (i = 1, 2, . . . , n) of A.
Form the matrix S whose column vectors
are the n basis eigenvectors of A.

Dr. Lê Xuân Đại (HCMUT-OISP) EIGENVALUES AND EIGENVECTORS HCMC — 2018. 24 / 56

 Diagonalization Definition

EXAMPLE

3.1 
15 −18 −16
Let A =  9 −12 −8  . Find a matrix S that
 
4 −4 −6
diagonalizes A.
Step 1. Find eigenvalues,
¯ eigenvectors of
¯ A.
¯ 15 − λ −18 −16 ¯¯
¯
χA (λ) = |A − λI| = ¯ 9 −12 − λ −8 ¯ = 0
¯ ¯
−6 − λ ¯
¯ ¯
¯ 4 −4
⇔ −(λ + 3)(λ + 2)(λ − 2) = 0 ⇔ λ1 = −3 (AM=1),
λ2 = −2 (AM=1), λ3 = 2 (AM=1).
Dr. Lê Xuân Đại (HCMUT-OISP) EIGENVALUES AND EIGENVECTORS HCMC — 2018. 25 / 56
 Diagonalization Definition

In the case where λ1 = −3 (AM=1),

  we have
 18x1 − 18x2 − 16x3 = 0
 1
9x1 − 9x2 − 8x3 = 0 ⇒ X1 = α  1  , α 6= 0.
 

 4x1 − 4x2 − 3x3 = 0 0

Dr. Lê Xuân Đại (HCMUT-OISP) EIGENVALUES AND EIGENVECTORS HCMC — 2018. 26 / 56

 Diagonalization Definition

In the case where λ2 = −2 (AM=1),

  we have
 17x1 − 18x2 − 16x3 = 0
 2
9x1 − 10x2 − 8x3 = 0 ⇒ X2 = β  1  , β 6= 0.
 

 4x1 − 4x2 − 4x3 = 0 1

Dr. Lê Xuân Đại (HCMUT-OISP) EIGENVALUES AND EIGENVECTORS HCMC — 2018. 27 / 56

 Diagonalization Definition

In the case where λ3 = 2 (AM=1),

  we have
 13x1 − 18x2 − 16x3 = 0
 4
9x1 − 14x2 − 8x3 = 0 ⇒ X3 = γ  2  , γ 6= 0.
 

 4x1 − 4x2 − 8x3 = 0 1

Dr. Lê Xuân Đại (HCMUT-OISP) EIGENVALUES AND EIGENVECTORS HCMC — 2018. 28 / 56

 Diagonalization Definition

Step
 2. Finda matrix S that diagonalizes A.
1 2 4
S= 1 1 2 
 
0 1 1
 
−3 0 0
Then S−1AS = D =  0 −2 0 
 
0 0 2

Dr. Lê Xuân Đại (HCMUT-OISP) EIGENVALUES AND EIGENVECTORS HCMC — 2018. 29 / 56

 Diagonalization Computing Powers of a Matrix

COMPUTING POWERS OF A MATRIX

Suppose A is diagonalizable, that is

S−1 AS = D = dig(λ1 , λ2 , . . . , λn ) ⇒ (S−1 AS)k = Dk ,
k ∈ N ⇒ S−1 A(S.S−1 )AS. . . . .S−1 AS = S−1 Ak S = Dk
⇒ Ak = SDk S−1 .
Therefore,
λk1
 
0 ... 0
 0 λk ... 0 
Ak = S  2  −1
S

 ... ... ... ... 
0 0 ... λkn
Dr. Lê Xuân Đại (HCMUT-OISP) EIGENVALUES AND EIGENVECTORS HCMC — 2018. 30 / 56
 Diagonalization Computing Powers of a Matrix

EXAMPLE

3.2 
0 −8 6
Let A =  −1 −8 7  . Find Ak , k ∈ N.
 
1 −14 11
¯ ¯
¯ −λ −8
¯ 6 ¯¯
χA (λ) = |A − λI| = ¯ −1 −8 − λ 7 ¯=0
¯ ¯
−14 11 − λ ¯
¯ ¯
¯ 1
⇔ −(λ − 2)(λ + 2)(λ − 3) = 0 ⇔ λ1 = −2 (AM=1),
λ2 = 2 (AM=1), λ3 = 3 (AM=1).

Dr. Lê Xuân Đại (HCMUT-OISP) EIGENVALUES AND EIGENVECTORS HCMC — 2018. 31 / 56

 Diagonalization Computing Powers of a Matrix

In the case where λ1 = −2 (AM=1),

   we have
 2x1 − 8x2 + 6x3 = 0
 1
−x1 − 6x2 + 7x3 = 0 ⇒ X1 = α  1  , α 6= 0.
 

 x − 14x + 13x = 0
1 2 3 1

Dr. Lê Xuân Đại (HCMUT-OISP) EIGENVALUES AND EIGENVECTORS HCMC — 2018. 32 / 56

 Diagonalization Computing Powers of a Matrix

In the case where λ2 = 2 (AM=1),

  we have
 −2x1 − 8x2 + 6x3 = 0
 1
−x1 − 10x2 + 7x3 = 0 ⇒ X2 = β  2  , β 6= 0.
 

 x − 14x + 9x = 0
1 2 3 3

Dr. Lê Xuân Đại (HCMUT-OISP) EIGENVALUES AND EIGENVECTORS HCMC — 2018. 33 / 56

 Diagonalization Computing Powers of a Matrix

In the case where λ3 = 3 (AM=1),

  we have
 −3x1 − 8x2 + 6x3 = 0
 2
−x1 − 11x2 + 7x3 = 0 ⇒ X3 = γ  3  , γ 6= 0.
 

 x − 14x + 8x = 0
1 2 3 5

Dr. Lê Xuân Đại (HCMUT-OISP) EIGENVALUES AND EIGENVECTORS HCMC — 2018. 34 / 56

 Diagonalization Computing Powers of a Matrix

A matrix
 S that
 diagonalizes
 A is 
1 1 2 1 1 −1
S =  1 2 3  ⇒ S−1 =  −2 3 −1  ,
   
1 3 5 1 −2 1
 
−2 0 0
D =  0 2 0  . Therefore, Ak = SDk S−1 =
 
0 0 3
   
1 1 2 (−2)k 0 0 1 1 −1
 1 2 3  0 2k 0   −2 3 −1 
   
1 3 5 0 0 3k 1 −2 1
Dr. Lê Xuân Đại (HCMUT-OISP) EIGENVALUES AND EIGENVECTORS HCMC — 2018. 35 / 56
 Diagonalization Recognizing Diagonalizability

THEOREM 3.1
The square matrix A is diagonalizable if and
only if the geometric multiplicity of every
eigenvalue is equal to the algebraic
multiplicity.

EXAMPLE

3.3 
2 0 1
Let A =  1 1 1  . Diagonalize A if A is
 
−2 0 −1
diagonalizable.
Dr. Lê Xuân Đại (HCMUT-OISP) EIGENVALUES AND EIGENVECTORS HCMC — 2018. 36 / 56
 Diagonalization Recognizing Diagonalizability

Step 1. Find eigenvalues, eigenvectors of A.

¯ ¯
¯ 2−λ 0 1 ¯
¯ ¯
χA (λ) = |A − λI| = ¯ 1 1 − λ 1 ¯=0
¯ ¯
0 −1 − λ
¯ ¯
¯ −2 ¯

⇔ −λ(λ − 1)2 = 0
⇔ λ1 = 0 (AM=1), λ2 = 1 (AM=2).

Dr. Lê Xuân Đại (HCMUT-OISP) EIGENVALUES AND EIGENVECTORS HCMC — 2018. 37 / 56

 Diagonalization Recognizing Diagonalizability

In the case where λ1 = 0 (AM=1),

  we have

 2x1 + x3 = 0 1
x1 + x2 + x3 = 0 ⇒ X1 = α  1  , α 6= 0.
 

 −2x − x = 0
1 3 −2

Dr. Lê Xuân Đại (HCMUT-OISP) EIGENVALUES AND EIGENVECTORS HCMC — 2018. 38 / 56

 Diagonalization Recognizing Diagonalizability

In the case where λ2 = 1 (AM=2), we have



 x1 + x3 = 0
x1 + x3 = 0

 −2x − 2x = 0
1
 3    
α 1 0
⇒ X2 =  β  = α  0  + β  1  , α + β2 6= 0.
      2
−α −1 0

Dr. Lê Xuân Đại (HCMUT-OISP) EIGENVALUES AND EIGENVECTORS HCMC — 2018. 39 / 56

 Diagonalization Recognizing Diagonalizability

Step
 2. The matrix
 that diagonalizes A is
1 0 1
S= 1 1 0 
 
−2 0 −1
 
−1 0 −1
−1
Then S =  1 1 1 
 
2 0 1
 
0 0 0
D = S−1 AS =  0 1 0 
 
0 0 1
Dr. Lê Xuân Đại (HCMUT-OISP) EIGENVALUES AND EIGENVECTORS HCMC — 2018. 40 / 56
 Diagonalization Recognizing Diagonalizability

EXAMPLE

3.4 
2 0 0
Let A =  0 4 0  . Diagonalize A if A is
 
1 0 2
diagonalizable.

Dr. Lê Xuân Đại (HCMUT-OISP) EIGENVALUES AND EIGENVECTORS HCMC — 2018. 41 / 56

 Diagonalization Recognizing Diagonalizability

Step 1. Find eigenvalues, eigenvectors of A.

¯ ¯
¯ 2−λ 0 0 ¯
¯ ¯
χA (λ) = |A − λI| = ¯ 0 4 − λ 0 ¯=0
¯ ¯
0 2−λ
¯ ¯
¯ 1 ¯

⇔ −(λ − 4)(λ − 2)2 = 0

⇔ λ1 = 4 (AM=1), λ2 = 2 (AM=2).

Dr. Lê Xuân Đại (HCMUT-OISP) EIGENVALUES AND EIGENVECTORS HCMC — 2018. 42 / 56

 Diagonalization Recognizing Diagonalizability

In the case where λ1 = 4(AM=1),

 we have
( 0
−2x1 = 0
⇒ X1 = α  1  , α 6= 0.
 
x1 − 2x3 = 0
0

Dr. Lê Xuân Đại (HCMUT-OISP) EIGENVALUES AND EIGENVECTORS HCMC — 2018. 43 / 56

 Diagonalization Recognizing Diagonalizability

In the case where λ2= 2(AM=2), we have

( 0
2x2 = 0
⇒ X2 = β  0  , β 6= 0.
 
x1 = 0
1
Step 2. Since the algebraic multiplicity =2>
geometric multiplicity=1 then A is not
diagonalizable.

Dr. Lê Xuân Đại (HCMUT-OISP) EIGENVALUES AND EIGENVECTORS HCMC — 2018. 44 / 56

 Diagonalization Orthogonal Diagonalization

ORTHOGONAL DIAGONALIZATION

DEFINITION 3.3
A square matrix A is said to be symmetric if
A = AT or equivalently if A = (aij )n then
aij = aji , ∀i, j = 1, 2, . . . , n.

THEOREM 3.2
If A is a symmetric matrix with real entries,
then the eigenvalues λ of A are all real
numbers and eigenvectors from different
eigenspaces are orthogonal.
Dr. Lê Xuân Đại (HCMUT-OISP) EIGENVALUES AND EIGENVECTORS HCMC — 2018. 45 / 56
 Diagonalization Orthogonal Diagonalization

DEFINITION 3.4
A square matrix P is said to be orthogonal if
its transpose is the same as its inverse, that is,
if PT = P−1, or equivalently, if PPT = PT P = I.

THEOREM 3.3
If A is a symmetric matrix with real entries,
then there exists the orthogonal matrix P
such that PT AP = P−1AP is diagonal.

Dr. Lê Xuân Đại (HCMUT-OISP) EIGENVALUES AND EIGENVECTORS HCMC — 2018. 46 / 56

 Diagonalization Orthogonal Diagonalization

ORTHOGONAL DIAGONALIZATION

Step 1. Find the eigenvalues.

Step 2. Find a basis for each eigenspace.
Step 3. Apply the Gram-Schmidt process to
each of these bases to obtain an
orthonormal basis for each eigenspace.
Step 4. Form the matrix P whose comlumns
are the vectors constructed in Step 3. The
eigenvalues on the diagonal of D = PT AP will
be the same order as their corresponding
eigenvectors in P.
Dr. Lê Xuân Đại (HCMUT-OISP) EIGENVALUES AND EIGENVECTORS HCMC — 2018. 47 / 56
 Diagonalization Orthogonal Diagonalization

EXAMPLE 3.5
Orthogonally
 diagonalize
 the matrix
2 −1 −1
A =  −1 2 −1 
 
−1 −1 2

Dr. Lê Xuân Đại (HCMUT-OISP) EIGENVALUES AND EIGENVECTORS HCMC — 2018. 48 / 56

 Diagonalization Orthogonal Diagonalization

Step 1. Find eigenvalues of A.

¯ ¯
¯ 2 − λ −1 −1 ¯
¯ ¯
χA (λ) = |A − λI| = ¯ −1 2 − λ −1 ¯=0
¯ ¯
−1 2 − λ
¯ ¯
¯ −1 ¯

⇔ −λ(λ − 3)2 = 0
⇔ λ1 = 0, (AM=1) λ2 = 3 (AM=2).

Dr. Lê Xuân Đại (HCMUT-OISP) EIGENVALUES AND EIGENVECTORS HCMC — 2018. 49 / 56

 Diagonalization Orthogonal Diagonalization

Step 2, 3.In the case where λ1 = 0 (AM=1),

2x1 − x2 − x3 = 0


we have −x1 + 2x2 − x3 = 0

 −x − x + 2x = 0
 1 2 3
1
⇒ X0 = α  1  , α 6= 0. Therefore,
 
1
 
p1
3
X0
p1
 
P∗1 = = 3

||X0 ||  
p1
3
Dr. Lê Xuân Đại (HCMUT-OISP) EIGENVALUES AND EIGENVECTORS HCMC — 2018. 50 / 56
 Diagonalization Orthogonal Diagonalization

In the case where λ2 = 3 (AM=2),

  we have
 −x1 − x2 − x3 = 0
 −α − β
−x1 − x2 − x3 = 0 ⇒ X =  α  =
 
β

 −x − x − x = 0
 1  2 3 
−1 −1
α  1  + β  0  , α2 + β2 6= 0.
   
0 1

Dr. Lê Xuân Đại (HCMUT-OISP) EIGENVALUES AND EIGENVECTORS HCMC — 2018. 51 / 56

 Diagonalization Orthogonal Diagonalization

Applying the Gram-Shmidt process, yields

the orthogonal
  eigenvectors B = {y1 , y2 }.
−1
y1 = X1 =  1  ,
 
0
 
−1/2
< X2 , y1 >
y2 = X2 − y1 =  −1/2 
 
< y1 , y1 >
1

Dr. Lê Xuân Đại (HCMUT-OISP) EIGENVALUES AND EIGENVECTORS HCMC — 2018. 52 / 56

 Diagonalization Orthogonal Diagonalization

Normalize the orthogonal basis to obtain

 
− p12 − p16
 
y1 y2  1 
P∗2 = = p1  and P∗3 = = − p6 
 
||y1 || 2 ||y2 ||  
2
0 p
6

Dr. Lê Xuân Đại (HCMUT-OISP) EIGENVALUES AND EIGENVECTORS HCMC — 2018. 53 / 56

 Diagonalization Orthogonal Diagonalization

Step 4. The matrix that

 orthogonally
p1 − p12 − p16
3
p1 p1 − p16 
 
diagonalizes A is P = 
 3 2 
p1 0 p2
 3  6
0 0 0
Then D = PT AP =  0 3 0 
 
0 0 3

Dr. Lê Xuân Đại (HCMUT-OISP) EIGENVALUES AND EIGENVECTORS HCMC — 2018. 54 / 56

 MatLab

MATL AB

1
Finding the characteristic polynomial
of A: p = poly(A)
2
Finding the roots of characteristic
equation of A: roots(p)
3
Finding eigenvalues and eigenvectors of
of A: [V , D] = eig(A)

Dr. Lê Xuân Đại (HCMUT-OISP) EIGENVALUES AND EIGENVECTORS HCMC — 2018. 55 / 56

 MatLab

THANK YOU FOR YOUR ATTENTION

Dr. Lê Xuân Đại (HCMUT-OISP) EIGENVALUES AND EIGENVECTORS HCMC — 2018. 56 / 56