MATH2033 (2023–2024)

Introduction to Scientific Computation

Linear Systems: Iterative Techniques III

PROBLEM SHEET SOLUTIONS

Solution to Problem 1
a We have that
     
2 −1 λ 0 2 − λ −1
− = .
−1 2 0 λ −1 2 − λ

Now,

2 − λ −1
= (2−λ)2 −(−1)2 = 4−4λ+λ2 −1 = λ2 −4λ+3 = (λ−1)(λ−3).
−1 2 − λ

So, the solutions to the characteristic equation

2 − λ −1
=0
−1 2 − λ

are λ = 1 and λ = 3. When λ = 1,

   
2 − λ −1 1 −1
=
−1 2 − λ −1 1
 
2 −1
and the eigenvectors of corresponding to the eigenvalue 1 are
−1 2
 
u1
the nonzero vectors that are such that
u2
    
1 −1 u1 0
= .
−1 1 u2 0

Hence, any nonzero vector of the form

 
α
α

Page 1 of 8
  
2 −1
is an eigenvector of corresponding to the eigenvalue 1. When
−1 2
λ = 3,    
2 − λ −1 −1 −1
=
−1 2 − λ −1 −1
 
2 −1
and the eigenvectors of corresponding to the eigenvalue 3 are
−1 2
 
v1
the nonzero vectors that are such that
v2
    
−1 −1 v1 0
= .
−1 −1 v2 0

Hence, any nonzero vector of the form

 
β
−β
 
2 −1
is an eigenvector of corresponding to the eigenvalue 3. There-
−1 2
fore, the eigenvalues of the matrix A are

• 1 withcorresponding
 eigenvectors that are the nonzero vectors of the
α
form ,
α
• 3 withcorresponding
 eigenvectors that are the nonzero vectors of the
β
form .
−β

b We have that
ρ(A) = max{|1|, |3|} = 3.

c The matrix A is not convergent since its spectral radius is not less than 1.

d We have that
      
T 2 −1 2 −1 4 + 1 −2 − 2 5 −4
A A= = = .
−1 2 −1 2 −2 − 2 1 + 4 −4 5

Now,      
5 −4 λ 0 5 − λ −4
− =
−4 5 0 λ −4 5 − λ
and

5 − λ −4
= (5−λ)2 −(−4)2 = 25−10λ+λ2 −16 = λ2 −10λ+9 = (λ−1)(λ−9).
−4 5 − λ

Page 2 of 8
 So the solutions to the characteristic equation

5 − λ −4
=0
−4 5 − λ

are λ = 1 and λ = 9. Consequently, the eigenvalues of AT A are 1 and 9.

So,
ρ AT A = max{|1|, |9|} = 9

and q
√
kAk2 = ρ AT A = 9 = 3.

Solution to Problem 2
a We have that
1 1
     
0 2
λ 0 −λ 2
1 − = 1 .
2
0 0 λ 2
−λ
Now,  2 
1
 
−λ 2
1 2 1 1
1 =λ − = λ− λ+ .
2
−λ 2 2 2
So, the solutions to the characteristic equation
1
−λ 2
1 =0
2
−λ

are λ = − 12 and λ = 12 . When λ = − 12 ,

1 1 1
   
−λ 2 2 2
1 = 1 1
2
−λ 2 2

1
 
0
and the eigenvectors of 1
2 corresponding to the eigenvalue − 12 are
0
 2
u1
the nonzero vectors that are such that
u2
1 1
    
2 2
u1 0
1 1 = .
2 2
u2 0
Hence, any nonzero vector of the form
 
α
−α
1
 
0
is an eigenvector of 1
2 corresponding to the eigenvalue − 12 . When
2
0
λ = 12 ,
1
− 12 1
   
−λ 2 2
1 = 1
2
−λ 2
− 12

Page 3 of 8
 1
 
0 2 1
and the eigenvectors of 1 corresponding to the eigenvalue 2
are
0
 2
v1
the nonzero vectors that are such that
v2

− 12 1
    
2
v1 0
1 = .
2
− 12 v2 0

Hence, any nonzero vector of the form

 
β
β

1
 
0
is an eigenvector of 1
2 corresponding to the eigenvalue 12 . There-
2
0
fore, the eigenvalues of the matrix A are

• − 12 with corresponding eigenvectors that are the nonzero vectors of

 
α
the form ,
−α
1
• 2
with corresponding eigenvectors that are the nonzero vectors of the
 
β
form .
β

b We have that  
1 1 1
ρ(A) = max − , = .
2 2 2

c The matrix A is convergent since ρ(A) < 1.

d We have that

1 1 1
    
T 0 2
0 2 4
0
A A= 1 1 = 1 .
2
0 2
0 0 4

Now,
1 1
     
4
0 λ 0 4
−λ 0
1 − = 1
0 4
0 λ 0 4
−λ
and 2
1

4
−λ 0 1
1 = −λ .
0 4
−λ 4
So the solution to the characteristic equation
1
4
−λ 0
1 =0
0 4
−λ

Page 4 of 8
 is λ = 14 . Consequently, the eigenvalue of AT A is 14 . So,

1 1
ρ AT A =


=
4 4
and r
1 1
q
T


kAk2 = ρ A A = = = 0.5.
4 2
Solution to Problem 3
a We have that

−1 2 0 −1 − λ
     
λ 0 0 2 0
 0 3 4 − 0 λ 0 = 0 3−λ 4 .
0 0 7 0 0 λ 0 0 7−λ
Now,

−1 − λ 2 0
3−λ 4
0 3−λ 4 = (−1−λ) = (−1−λ)(3−λ)(7−λ).
0 7−λ
0 0 7−λ
So, the solutions to the characteristic equation

−1 − λ 2 0
0 3−λ 4 =0
0 0 7−λ
are λ = −1, λ = 3 and λ = 7. When λ = −1,

−1 − λ
   
2 0 0 2 0
 0 3−λ 4 = 0 4 4 
0 0 7−λ 0 0 8

−1
 
2 0
and the eigenvectors of  0 3 4  corresponding to the eigenvalue −1
0 0 7
 
u1
are the nonzero vectors u2
  that are such that
u3
    
0 2 0 u1 0
 0 4 4   u2  =  0  .
0 0 8 u3 0
Hence, any nonzero vector of the form
 
α
 0 
0

Page 5 of 8
 −1 2 0
 

is an eigenvector of  0 3 4  corresponding to the eigenvalue −1.

0 0 7
When λ = 3,

−1 − λ −4
   
2 0 2 0
 0 3−λ 4 = 0 0 4 
0 0 7−λ 0 0 4

−1 2 0
 

and the eigenvectors of  0 3 4  corresponding to the eigenvalue 3

0 0 7
 
v1
are the nonzero vectors  v2  that are such that
v3

−4 2 0
    
v1 0
 0 0 4   v2  =  0  .
0 0 4 v3 0

Hence, any nonzero vector of the form

 
β
 2β 
0

−1 2 0
 

is an eigenvector of  0 3 4  corresponding to the eigenvalue 3.

0 0 7
When λ = 7,

−1 − λ −8
   
2 0 2 0
 0 3−λ 4  =  0 −4 4 
0 0 7−λ 0 0 0

−1 2 0
 

and the eigenvectors of  0 3 4  corresponding to the eigenvalue 7

0 0 7
 
w1
are the nonzero vectors  w2  that are such that
w3

−8
    
2 0 w1 0
 0 −4 4   w2  =  0  .
0 0 0 w3 0

Page 6 of 8
 Hence, any nonzero vector of the form
 
γ
 4γ 
4γ
−1 2 0
 

is an eigenvector of  0 3 4  corresponding to the eigenvalue 7.

0 0 7
Therefore, the eigenvalues of the matrix A are

• −1 with corresponding eigenvectors that are the nonzero vectors of

 
α
the form  0 ,
0
• 3 with corresponding eigenvectors that are the nonzero vectors of the
 
β
form  2β ,
0
• 7 with corresponding eigenvectors that are the nonzero vectors of the
 
γ
form  4γ .
4γ

b We have that
ρ(A) = max{| − 1|, |3|, |7|} = 7.

c The matrix A is not convergent since its spectral radius is not less than 1.
Solution to Problem 4
If λ is an eigenvalue of A and v is an eigenvector of A corresponding to λ then

AT Av = AT (λv) = λAT v = λAv = λ2 v

since AT = A as A is symmetric. Consequently, the eigenvalues of AT A are
λ21 , λ22 , . . . , λ2n where λ1 , λ2 , . . . , λn are the eigenvalues of A. Furthermore,
q
kAk2 = ρ(AT A)
r
= max |λ2j |
j∈{1,2,...,n}
r
= max (|λj |2 )
j∈{1,2,...,n}
s 2
= max |λj |
j∈{1,2,...,n}

= max |λj |
j∈{1,2,...,n}

= ρ(A).

Page 7 of 8
References

• Burden, Faires & Burden, Numerical Analysis, 10E

– Section 7.2

Page 8 of 8