MATH3060 - M ATHEMATICAL A NALYSIS III - T UTORIAL 7

TA: LEE, Yat Long Luca

Email: [email protected]
Office: Room 711 AB1 (Temporary), Room 505 AB1 (Until further notice)
Office Hour: Send me an email first, then we will arrange a meeting (if you need it).

Contraction Mapping Principle

Recall from the previous lectures, we learnt about:

Definition 1
A map T : ( X, d) → ( X, d) is a contraction if there exists a constant γ ∈ (0, 1) such that

d( Tx, Ty) ≤ γd( x, y)

for all x, y ∈ X.

Theorem 1 (Contraction mapping principle)

Every contraction in a complete metric space admits a unique fixed point.

In this tutorial, we prove the following corollary:

Corollary 1 (Source: Functional Analysis by S. Kesavan P.55)

Let ( X, d) be a complete metric space and let T : ( X, d) → ( X, d) be a map such that for some positive
integer n, the map T n = T ◦ · · · ◦ T : ( X, d) → ( X, d) is a contraction. Then T has a unique fixed
point.

Proof of Corollary 1:
Denote the fixed point of T n by x∗ , then

Tx∗ = TT n x∗ = T n+1 x∗ = T n Tx∗

that means Tx∗ is also a fixed point of T n . However the contraction mapping principle tells us
that the fixed point x∗ is unique. This implies Tx∗ = x∗ .

Uniqueness follows from the fact that any fixed point x∗ of T is also a fixed point of T n ,
because
T n x∗ = T n−1 Tx∗ = T n−1 x∗ = · · · = Tx∗ = x∗

Thus T has unique fixed point.

!

1
 MATH3060 - M ATHEMATICAL A NALYSIS III - T UTORIAL 7

Exercise 1
Show that every continuous function f : [0, 1] → [0, 1] has a fixed point.

Solution:

If f (0) = 0 or f (1) = 1, then we are done. Hence, we assume f (0) > 0 and f (1) < 1.
Define g : [0, 1] → R by g( x) = f ( x) − x. Then g(0) = f (0) > 0 and g(1) = f (1) − 1 < 0.

Since f is continuous, then g is continuous. By the intermediate value theorem, there exists
a c ∈ [0, 1] such that
g(c) = f (c) − c = 0 =⇒ f (c) = c

thus, f has a fixed point

!

2
 MATH3060 - M ATHEMATICAL A NALYSIS III - T UTORIAL 7

Exercise 2
(Source: Previous HW problem of MATH3060)
Fix α ∈ [0, 1), for each x0 ∈ [0, 1], consider the iteration sequence

xn = αxn−1 (1 − xn−1 ), ∀n ∈ N

(a) Show that { xn } ⊆ [0, 1]

(b) Show that lim xn = 0

n→∞

Solution:

We first show that { xn } ⊆ [0, 1]. Define T : [0, 1] → R, by Tx = αx(1 − x). Then the iter-
ation sequence can be written as
xn = Txn−1

Then it is equivalent to show that T ([0, 1]) ⊂ [0, 1].

Since T is smooth, we differentiate T and get T ′ ( x) = α (1 − 2x) which implies that T at-
tains its maximum at x = 12 , and the value is T ( 12 ) = α4 < 1 since α < 1. Moreover, it is obvious
that Tx ≥ 0 for all x ∈ [0, 1]. Thus, T ([0, 1]) ⊂ [0, 1]. It follows that xn = T n x0 ∈ [0, 1] for all n.

First of all we know that T (0) = 0. Then we check that T is a contraction

| Tx − Ty| = | T ′ (c)|| x − y| ≤ M| x − y|

where M := max | T ′ ( x)|, which can be calculated to be α < 1. Hence for all x, y ∈ [0, 1], T is a
x∈[0,1]
contraction. Then the contraction mapping principle tell us that the fixed point x = 0 is unique.

Now suppose that xn → L as n → ∞, then

L = lim xn = T ( lim xn−1 ) = T ( L) =⇒ L = 0

n→∞ n→∞

by the contraction mapping principle.

!

3
 MATH3060 - M ATHEMATICAL A NALYSIS III - T UTORIAL 7

Exercise 3
(Source: Ordinary Differential Equations, Lecture Notes of MATH4051 at HKUST, by Prof
Frederick Fong.)

Define g( x) = cos x − 13 cos3 x where x ∈ [0, π3 ]. Consider the iteration sequence

!
"x = 0
0
, ∀n ≥ 1
# xn = g ( xn−1 )

Show that g( x) maps [0, π3 ] to [0, π3 ] and that it satisfies the contraction inequality. Hence,
show that the iteration sequence xn converges to a limit L which is the root of the equation
x = cos x − 13 cos3 x.

Solution:
Since g( x) is smooth, we can consider its derivative:

g′ ( x) = − sin x − cos2 x(− sin x)

= − sin x(1 − cos2 x)
= − sin3 x

then one can see that g′ ( x) ≤ 0 for all x ∈ [0, π3 ], hence it is decreasing.
Note that g(0) = 1 − 13 = 23 < π3 , and that g( π3 ) = 12 − 24 1
= 11
24 > 0. Together with g the
fact that is decreasing, we see that g([0, 3 ]) ⊂ [0, 3 ].
π π

Then for all x, y ∈ [0, π3 ], by the mean value theorem, one can see that

| g( x) − g( y)| = | g′ (c)|| x − y| ≤ M| x − y|
√
3 3
where M := maxπ | g′ ( x)|. On [0, π3 ], we have M = | sin3 ( π3 )| = 8 <1
x∈[0, 3 ]

Now consider the iteration sequence xn = g( xn−1 ) , we have

| xn+1 − xn | = | g( xn ) − g( xn−1 )|
≤ M | xn − xn−1 |
≤ M 2 | xn−1 − xn−2 |
≤ M n−1 | x2 − x1 |

Then we see that

∞ ∞
∑ | xn+1 − xn | ≤ ∑ M n−1 | x2 − x1 |
n=0 n=0

converges, since the RHS is a geometric series and M < 1. Then the series
∞
∑ ( xn+1 − xn )
n=0

converges absolutely. Hence x N := x1 + ∑nN=−11 ( xn+1 − xn ) converges as N → ∞.

4
 MATH3060 - M ATHEMATICAL A NALYSIS III - T UTORIAL 7

Suppose that xn → L as n → ∞, then

L = lim xn = g( lim xn ) = g( L)
n→∞ n→∞

implies that L is a root of the equation x = g( x).

!

Possible Reference
Instead of giving you another exercise, the following books contain many examples in which
you can take a look at them if you are interested:

• Real Analysis by Royden and Fitzpatrick

• Metric Spaces by Copson