MATH2033 (2023–2024)
Introduction to Scientific Computation
Nonlinear Equations II
PROBLEM SHEET SOLUTIONS
Solution to Problem 1
a We first note that g is real-valued and continuous on [0, 2π]. Moreover,
since 0 ≤ sin x2 ≤ 1 for all x ∈ [0, 2π], π ≤ g(x) ≤ π + 12 for all
x ∈ [0, 2π]. Hence, g(x) ∈ [0, 2π] for all x ∈ [0, 2π]. Furthermore, g 0 (x) =
1 x
and so g is differentiable on (0, 2π) and |g 0 (x)| ≤ 14 < 1 for all
4
cos 2
x ∈ (0, 2π). Hence, by the theorem on page 5 of the Nonlinear Equations
II slides, g has a unique fixed point in [0, 2π].
b With k = 14 , p0 = π , a = 0 and b = 2π , the theorem on page 14 of the
Nonlinear Equations II slides yields that
n
n 1
|p − pn | ≤ k max{π − 0, 2π − π} = π.
4
We want |p − pn | ≤ 10−2 . This will be satisfied when π
4n
≤ 10−2 . Now,
π
≤ 10−2
4n
⇔ 4n ≥ 102 π
⇔ log10 (4n ) ≥ log10 (100π)
⇔ n log10 (4) ≥ log10 (100π)
log10 (100π)
⇔ n≥ .
log10 (4)
log (100π)
Moreover, log 10
= 4.147 . . . and so |p − pn | ≤ 10−2 if at least 5 itera-
10 (4)
tions are performed.
c Starting with p0 = π and using
1 p
n−1
pn = g(pn−1 ) = π + sin
2 2
yields
p1 = 3.641592 . . . ,
p2 = 3.626048 . . . ,
Page 1 of 2
p3 = 3.626995 . . . ,
p4 = 3.626938 . . .
and
p5 = 3.626942 . . . .
Solution to Problem 2
a Suppose that pn → p as n → ∞ and that p 6= 0. Then
p = lim pn = lim g(pn−1 ) = g lim pn−1 = g(p)
n→∞ n→∞ n→∞
since lim pn−1 = p and g is continuous. Therefore,
n→∞
p = g(p).
Now, since p 6= 0,
1
p = g(p) ⇔ p = 2p−Ap2 ⇔ p−Ap2 = 0 ⇔ p(1−Ap) = 0 ⇔ 1−Ap = 0 ⇔ p = .
A
So, if the sequence {pn }∞
n=0 converges to a nonzero limit then that limit is
1
A
.
b We first note that g is continuous and differentiable on all intervals about
1
A
. Now, g 0 (x) = 2 − 2Ax and so g 0 is a linear polynomial, g 0 (x) = −1 ⇔
2 − 2Ax = −1 ⇔ 2Ax = 3 ⇔ x = 2A 3
and g 0 (x) = 1 ⇔ 2 − 2Ax =
1
. Hence, −1 < g 0 (x) < 1 for all x ∈ 2A 1 3
1 ⇔ 2Ax = 1 ⇔ x = 2A , 2A .
0 1 3
Therefore, |g (x)| ≤ k , with 0 < k < 1, for all x ∈ 2A + c, 2A − c where
1 1
1 3
c ∈ 0, 2A . Let us choose c = 4A in which case 2A + c, 2A −c =
0
3 5 1 1
1 3
3
,
4A 4A
. Now, g (x) = 0 ⇔ x = A
and g A
= A
, g 4A
= 2 4A −
9 3 9 24 9 15 5 5 25
A 16A2 = 2A − 16A = 16A − 16A = 16A and g 4A = 2 4A − A 16A2 =
5 25 40 25 15
3 5 15 1
2A
− 16A
= −
16A 16A
= 16A
. Consequently, g maps , onto ,A .
15 1
3 5 3 5 4A 4A 16A
3 5
Therefore, since 16A , A ⊂ 4A , 4A , g(x) ∈ 4A , 4A for all x ∈ 4A , 4A .
Hence, by the theorem on page 12 of the Nonlinear Equations II slides
with a = 4A 3
and b = 4A 5
, the sequence {pn }∞ n=0 converges for any p0 ∈
3 5
,
4A 4A
.
Solution to Problem 3
Let g be given by g(x) = 1 − x2 . Then g is continuous and real-valued on [0, 1]
and g(x) ∈ [0, 1] for all x ∈ [0, 1]. Moreover, g is differentiable on (0, 1) and
g 0 (x) = −2x ≤ k , with 0 < k < 1, for all x ∈ (0, 1). However, starting with
p0 = 1 ∈ [0, 1] and using pn = g(pn−1 ) yields p1 = g(p0 ) = 1 − 12 = 0 and
1 if n is even,
p2 = g(p1 ) = 1−02 = 1 = p0 . So pn = Hence, the sequence
0 if n is odd.
{pn }∞ n=0 does not converge.
References
• Burden, Faires & Burden, Numerical Analysis, 10E
– Section 2.2
Page 2 of 2
