MATH2033 (2023–2024)

Introduction to Scientific Computation

Linear Systems: Iterative Techniques I

PROBLEM SHEET SOLUTIONS

Solution to Problem 1
a The approximations obtained after k iterations of the Jacobi method are
given by
(k)1 (k−1) (k−1)

x1 = 1 + x2 − x3 ,
3
(k) 1 (k−1) (k−1)

x2 = −3x1 − 2x3
6
and
(k) 1 (k−1) (k−1)

x3 = 4 − 3x1 − 3x2 .
7
(0) (0) (0)
Starting with x1 = x2 = x3 = 0, we can compute that
(1) 1
x1 = = 0.3333333333 . . . ,
3
(1)
x2 = 0
and
(1) 4
x3 = = 0.5714285714 . . . .
7
So  
0.3333333333 . . .
x(1) = 0 .
0.5714285714 . . .
We can then compute that
(2) 1
x1 = (1 − 0.5714281754 . . .) = 0.1428571428 . . . ,
3
(2) 1
x2 = (−1 − 1.1428571428) = −0.3571428571 . . .
6
and
(2) 1
x3 = (4 − 1) = 0.4285714285 . . . .
7
So  
0.1428571428 . . .
x(2) =  −0.3571428571 . . .  .
0.4285714285 . . .

Page 1 of 5
 b The approximations obtained after k iterations of the Gauss–Seidel method
are given by
(k) 1 (k−1) (k−1)

x1 = 1 + x2 − x3 ,
3
(k) 1 (k) (k−1)

x2 = −3x1 − 2x3
6
and
(k) 1 (k) (k)

x3 = 4 − 3x1 − 3x2 .
7
(0) (0) (0)
Starting with x1 = x2 = x3 = 0, we can compute that

(1) 1
x1 = = 0.3333333333 . . . ,
3
(1) −1
x2 = = −0.1666666666 . . .
6
and
(1) 4 − 1 + 0.5
x3 = = 0.5.
7
So  
0.3333333333 . . .
x(1) =  −0.1666666666 . . .  .
0.5
We can then compute that

(2) 1
x1 = (1 − 0.1666666666 . . . − 0.5) = 0.1111111111 . . . ,
3
(2) 1
x2 = (−0.3333333333 . . . − 1) = −0.2222222222 . . .
6
and
(2) 1
x3 = (4 − 0.3333333333 . . . + 0.6666666666 . . .) = 0.6190476190 . . . .
7
So  
0.1111111111 . . .
x(2) =  −0.2222222222 . . .  .
0.6190476190 . . .

Solution to Problem 2
a The approximations obtained after k iterations of the Jacobi method are
given by
(k) 1  (k−1)

x1 = 9 + x2 ,
10
(k) 1  (k−1) (k−1)

x2 = 7 + x1 + 2x3
10
and
(k) 1  (k−1)

x3 = 6 + 2x2 .
10
Page 2 of 5
 (0) (0) (0)
Starting with x1 = x2 = x3 = 0, we can compute that

(1) 9
x1 = = 0.9,
10
(1) 7
x2 = = 0.7
10
and
(1) 6
x3 = = 0.6.
10
So  
0.9
x(1) =  0.7  .
0.6
We can then compute that

(2) 1
x1 = (9 + 0.7) = 0.97,
10
(2) 1
x2 = (7 + 0.9 + 1.2) = 0.91
10
and
(2) 1
x3 = (6 + 1.4) = 0.74.
10
So  
0.97
x(1) =  0.91  .
0.74

b The approximations obtained after k iterations of the Gauss–Seidel method

are given by
(k) 1  (k−1)

x1 = 9 + x2 ,
10
(k) 1  (k) (k−1)

x2 = 7 + x1 + 2x3
10
and
(k) 1  (k)

x3 = 6 + 2x2 .
10
(0) (0) (0)
Starting with x1 = x2 = x3 = 0, we can compute that

(1) 9
x1 = = 0.9,
10
(1) 7 + 0.9
x2 = = 0.79
10
and
(1) 6 + 1.58
x3 = = 0.758.
10

Page 3 of 5
 So  
0.9
x(1) =  0.79  .
0.758
We can then compute that

(2) 1
x1 = (9 + 0.79) = 0.979,
10
(2) 1
x2 = (7 + 0.979 + 1.516) = 0.9495
10
and
(2) 1
x3 = (6 + 1.899) = 0.7899.
10
So  
0.979
x(2) =  0.9495  .
0.7899

Solution to Problem 3
a As A is positive definite, xT Ax > 0 for every nonzero x ∈ Rn . For
i = 1, 2, . . . , n, let xi ∈ Rn be the vector with entries

1 if j = i,
xij =
0 if j 6= i,

for j = 1, 2, . . . , n. Then, for i = 1, 2, . . . , n,

 
a1i
 a2i 

T 

T 
xi Axi = xi  .  = aii
 .. 
ani
T
and (xi ) Axi > 0 as A is positive definite. Therefore, all of the entries on
the main diagonal of A are positive.

b We can write A = D − L − U , where D is a diagonal matrix, L is a strictly

lower triangular matrix and U is a strictly upper triangular matrix. Now, the
entries on the main diagonal of D are all positive because the entries on
the main diagonal of A are all positive. Moreover, since A is symmetric,
D − L − U = A = AT = D T − LT − U T = D − LT − U T and so
L + U = LT + U T from which we can conclude that U = LT because
LT is a strictly upper triangular matrix and U T is a strictly lower triangular
matrix. Therefore, we can write A = D − L − LT , where D is a diagonal
matrix where the entries on the main diagonal are all positive and L is a
strictly lower triangular matrix.

Page 4 of 5
 c Since the entries on the main diagonal of D are all positive, the entries on
the main diagonal of D − L are all positive. So, since D − L is a lower
triangular matrix, det(D − L) 6= 0. Hence, D − L is nonsingular.

d Since A = D − L − LT , LT = D − L − A and so

M = (D − L)−1 LT
= (D − L)−1 (D − L − A)
= (D − L)−1 (D − L) − (D − L)−1 A
= I − (D − L)−1 A.

References

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

– Section 7.3

Page 5 of 5