1 EIGENVALUES AND EIGEN-

VECTORS
Objectives
By the end of this topic, you will be able to:

ˆ compute eigenvalues and eigenvectors of square matrices.

ˆ diagonalize matrices.

ˆ perform orthogonal diagonalization of symmetric matrices.

Main Reference:Schaum’s outline Linear Algebra 4th edition, chapter 9.

1.1 Eigenvalues and Eigenvectors

Definition:
Let A be a square matrix. The real number λ is called eigenvalue of A if ∃
a nonzero vector x such that,

Ax = λx

x is called the eigenvector of A.

Examples: ! !
3 0 1
#1. Let A = ; x= & λ = 3.
8 −1 2
Show that Ax = λx.
! !
4 −2 2
#2. Let A = & x= .
1 1 1

Find the eigenvalue of A.

1.2 Real Life Applications

i) System of Communication: Eigenvalues are used to calculate the theo-
retical limit of how much information can be carried via a communica-

1
 tion channel such as a telephone line or the air. The eigenvectors and
eigenvalues of the communication channel (represented as a matrix)
are calculated, and then the eigenvalues are water-filled. The eigenval-
ues are then essentially the gains of the channel’s fundamental modes,
which are recorded by the eigenvectors.

ii) Bridge Construction: The smallest magnitude eigenvalue of a system

that models the bridge is the natural frequency of the bridge. Engineers
use this knowledge to guarantee that their structures are stable.

iii) Automobile Stereo System Design: Eigenvalue analysis is also employed

in the design of car stereo systems, where it aids in the reproduction
of car vibration caused by music.

iv) Electrical and Mechanical Engineering: The use of eigenvalues and

eigenvectors to decouple three-phase systems via symmetrical compo-
nent transformation is advantageous.

v) In Statistics: Computing CIs, PCA and Factor Analyses, Matrix Fac-

torization, LS Approximations etc.

1.3 Algebraic Eigenvalue Problem

To find the eigenvalues of a square matrix A, we rewrite the equation Ax = λx
as Ax = λIx.

=⇒ (λI − A)x = 0 OR (A − λI) = 0

This system has a non trivial solution iff

det(λI − A) = 0 OR det(A − λI) = 0

The equation det(λI − A) = 0 is called the characteristic equation of A and

it is expanded, its called the characteristic polynomial of A.
Examples:
#3. Find the eigenvalues of:

2
 ! !
3 0 3 2
i) A = ii) B = .
8 −1 −1 0
Hint: Solve | λI − A |= 0 and | λI − B |= 0
#4. For each eigenvalue above, find the associated eigenvectors of the ma-
trices.
#5. Find the eigenvalues and eigenvectors (for integer eigenvalues) of the
following
 matrices     
1 2 −1 0 1 0 5 6 2
     
i) A = 1 0
 1 ii) B = 0 0 1

 iii) C = 0 −1 −8

.
4 −4 5 4 −17 8 1 0 −2
Theorem:
If A is square matrix, the following are equivalent:

a) λ is the eigenvalue of A

b) (λI − A)x = 0 has nontrivial solutions

c) ∃ a nonzero vector x ∈ Rn such that Ax = λx

d) λ is a real solution to | λI − A |= 0.

The eigenvectors are the nonzero vectors in the solution space of (λI −A)x =
0. They are called eigenspace of A corresponding to
λ. 
3 −2 0
 
#6. Compute the bases for the eigenspace of A =   −2 3 0 .

0 0 5
Losung:
2
X-teristic eqn: | λI − A |= 0=⇒ (λ − 1)(λ
 −5) =0; 
∴ λ = 1, 5. Verify.
−s −1 0
     
For λ = 5: eigenvector x =   s  = s  1  + t 0.Work these out!
    
t 0 1
   
−1 0
   
Since 
 1  and 0 are LI, they form a basis for the eigenspace corre-
  
0 1
sponding to λ = 5.

3
    
t 1
   
Similarly for λ = 1 the eigenvector x = 
 t  = t 1.
  
0 0
 
1
 
Hence 1

 forms a basis corresponding to λ = 1.
0
EER: Check out relevant questions from Exercise Set 5.1 from the main
text above.

1.4 Diagonalization
Definition:
A matrix A is said to be diagonalizable if ∃ an invertible matrix P such that
P −1 AP is diagonal, the matrix!P is said to diagonalize
! A.
1 1 1 1
For example, if A = ; P = . Then D = P −1 AP =
−2 4 1 2
!
2 0
. Check it out!
0 3
Theorem:
If A is n × n and diagonalizable, then A has n LI eigenvectors.

1.4.1 Procedure for Diagonalizing Matrices

1) Find n LI eigenvectors of A, e.g {v1 , v2 , · · · , vn }

2) Form the matrix P having v1 , v2 , · · · , vn as its column vectors.

3) The matrix P −1 AP will then be diagonal with λ1 , λ2 , · · · , λn as its di-

agonal elements where λi is the eigenvalue corresponding to the eigen-
vector vi ∀ i
 
3 −2 0
 
#7. Find the matrix P that diagonalizes A = 
−2 3 .
0
0 0 5
Losung:

4
From an earlier example (#6 above), the eigenvalues ofA are λ =1, 5 and
1 0 −1
 
the eigenvectors form the column vectors of P, i.e P = 
1 0 1  or any
0 1 0
similar arrangement of the eigenvectors of A.
Verify that D = P −1 AP is diagonal.
#8. EER: Find the matrix P that diagonalizes the matrices in example
5 (#5) above (for integer eigenvalues) and verify that the corresponding
P −1 AP are diagonal. !
−3 2
#9. Find the characteristic equation of A = and show that A is
−2 1
not diagonalizable.
Losung: Verify! | λI − A |= 0 =⇒ (λ + 1)2 = 0 =⇒ λ = −1.
Only one eigenvalue implying only one eigenvector (not two).
Hence A is not diagonalizable.  
x1
 
#10. Let T : R3 3
→ R be a linear operator given by T 
x2  =
 
x3
 
3x1 − 2x2
 
−2x1 + 3x2 .
 
5x3
Find a basis for R3 relative to which the matrix  of T
 is diagonal. 
3 −2
   
Losung: Basis for R3 = (e1 , e2 , e3 ) ⇒ T (e1 ) =  −2
 
 ; T (e2 ) =  3  & T (e3 ) =
 
0 0
 
0
 
0.
 
5
 
3 −2 0
 
Thus the standard matrix A for T is = −2 3 0

.
0 0 5
Change to a new basis, B = (u1 , u2 , u3 ) to obtain the diagonal matrix A′ for
′ ′ ′

5
T . Ifthe transition
 matrix P diagonalizes
 A, then A′ = P −1 AP . From #7,
−1 0 1 5 0 0
  ′
 
P =  1 0 1  & A = 0 5 0.
  
0 1 0 0 0 1
The columns of P are [u′1 ]B , [u′2 ]B ; [u′3 ]B .
Thus, 
−1
u′1 = (−1)e1 + (0)e2 + (0)e3 = 
 
1

0
 
0
′
 
u2 = (0)e1 + (0)e2 + (1)e3 = 0


1
 
1
′
 
u3 = (1)e1 + (1)e2 + (0)e3 = 1


0
are the basis vectors that produce the diagonal matrix A′ .
Theorems:

a) If v1 , v2 , · · · , vn are eigenvectors corresponding to the distinct eigen-

values λ1 , λ2 , · · · , λn , then {v1 , v2 , · · · , vn } is a linearly independent
set.

b) If an n×n matrix A has n distinct eigenvalues, then A is diagonalizable.

EER: 

−1 0 1
 
#11. Show that the matrix 
 −1 3 0 is not diagonalzable.

0 0 3
−1
#12. Find the matrix
 P that diagonalizes A and hence fine P AP where
2 0 −2
 
A=  0 3 0 .

−4 13 −1

6
  
x1
  
#13. Let T : R3 → R3 be a linear operator given by T  x2  =
 
x3
 
2x1 − x2 − x3
 
 x1 − x3 .
 
−x1 + x2 + 2x3
Find a basis for R3 relative to which the matrix of T is diagonal.

1.5 Orthogonal Diagonalization

For a given linear operator; L : V → V , we wish to find an orthonor-
mal basis for V for which the matrix of L is diagonal. OR Alternatively,
given a square matrix A, we wish to find an orthogonal matrix P such
that P −1 AP = (P T AP ) is diagonal. (Note that for orthogonal matrix B,
B −1 = B T or BB T = I).
Definitions:

ˆ A square matrix A is called orthogonally diagonalizable if there is an

orthogonal matrix P s.t. P −1 AP (= P T AP ) is diagonal; the matrix P
is said to orthogonally diagonalze A.

ˆ A square matrix A is symmetric if A = AT .

For symmetric matrices, off diagonal elements are the ”same”. Give some
examples....
Theorem:
If A is symmetric, then eigenvectors from different eigenspaces are orthogo-
nal.

1.5.1 Procedure for Orthogonally Diagonalizing Symmetric Ma-

trix

1) Find a basis for each eigenspace of the matrix A.

2) Apply the Gram-Schmidt process to each basis to obtain an orthonor-

7
 mal basis for each eigenspae.

3) Form the matrix P whose columns are the orthonormal basis. P or-
thogonally diagonalizes A.
 
4 2 2
 
#14. Find an orthogonal matrix P that diagonalizes A = 
 2 4 2.

2 2 4
Losung:
Find eigenvalues of A, i.e | 
λI − (λ −2)2 (λ − 8) = 0. ∴ λ = 2, 8.
A |= 0 =⇒
−1 −1
   
Eigenvectors: λ = 2, u1 =   1  ; u2 =  0 . Verify! u1 & u2 form a
  
0 1
basis.
Applying Gram-Schmidt process to u1 and u2 to obtain orthonormal bases
v1 and v2 :    
−1
−1 √
u1 1    12 
  
v1 = = √ 1 = √ 
||u1 || 2    2
0 0
 
−1
√
u2 − < u2 , v1 > v1  6
−1 
v2 = = √
|| u2 − < u2 , v1 > v1 ||  6 
√2
6
 
−1
 2
Note that < u2 , v1 >= u2 .v1 and u2 − < u2 , v1 > v1 = − 1 ; verify.
 2
0
 
1
 
For λ = 8, the basis is u3 = 
1 (verify.

1
Gram-Schmidt on u3 :  
1
u3 1  
v3 = =√  1
||u3 || 3 
1

8
The matrix P is made up columns of v1 , v2 , v3 i.e
 
−1
√ −1
√ √1
 2 6 3
1 −1 √1  ⊡
P =
 √2 √
6 3
0 √2 √1
6 3

EER: Verify that P T AP is diagonal.

Theorems:

a) The characteristic equation of a symmetric matrix A has only real roots.

b) If an eigenvalue λ of a symmetric matrix A is repeated k times as a

root to the characteristic equation, then the eigenspace corresponding
to λ is k− dimensional.

EER:
#15. Find the dimensions of the eigenspaces of the following symmetric
matrices:  
10
  − 43 0 − 43
1 1 1  34
− 53 1

  − 3 0 3

i) A =  1 1 1; λ = 0(2D), 3(1D) ii) B = 
 ;
 
 0 0 −2 0 
1 1 1

− 34 1
3
5
0 −3
λ = −2(3D), 4(1D)

#16. Find a matrix P which orthogonally diagonalizes A and determine

P T AP : ! !
√1 −1
3 1 √
i) A = ; P= 2 2
; P T AP =??
1 3 √1 √1
2 2
   
1 1 0 √1 √1 0
   2 2 
ii) A = 1 1 0  √1 −1 P T AP =??
; P= √
;
0

 2 2
0 0 0 0 0 1

9
    
3 1 0 0 0 0 √1 √1
2 2
√1 −1 
   
1 3 0 0 0 0 √
iii) A = 
 ; P= 2 2 ; P T AP =??
 0 0 0 0
 1 0

0 0


0 0 0 0 0 1 0 0

GOOD LUCK in your end of Semester Exams

10