Sunteți pe pagina 1din 33

Metric Spaces

Lecture Notes and Exercises, Fall 2015

M.van den Berg

School of Mathematics University of Bristol BS8 1TW Bristol, UK mamvdb@bristol.ac.uk

1

Deﬁnition of a metric space.

Let X be a set, X

= . Elements of X will be called points.

Deﬁnition 1.1 (a) A map d : X × X R is called a metric on X if

 (M1) (∀(x, y) ∈ X × X) [0 ≤ d(x, y) < ∞] . (M2) (∀(x, y) ∈ X × X) [(d(x, y) = 0) ⇔ (x = y)] . (M3) (∀(x, y) ∈ X × X) [d(x, y) = d(y, x)] . (M4) (∀(x, y, z) ∈ X × X × X) [d(x, y) ≤ d(x, z) + d(z, y)] .

(b) (X, d) is called a metric space and d(x, y) is the distance between points x and y.

Remark 1.2 Conditions (M1)-(M4) are the axioms of a metric space. (M4) is called the triangle inequality.

Example 1.3 Let X be a non-empty set. Deﬁne d : X 2 R by

d(x, y) =

1,

0,

if

if

x

x = y.

= y,

Then d is a metric on X, and (X, d) is a discrete metric space.

Example 1.4 Let R be the real line with d(x, y) = |x y|. Then (R, d) is a metric space.

 Example 1.5 Let X = R N := {x = (x 1 , x 2 , deﬁned by  N d(x, y) =   j=1

Then (R N , d) is a metric space.

, x N ) | x i R,

(x j y j ) 2

1/2

1 i N }.

.

Let d : X 2 R be

Theorem 1.6 Let X = l 2 := {x = (x n ) nN |

n=1 x

n 2 < ∞}. Let d : X 2 R be deﬁned

d(x, y) =

Then (l 2 , d) is a metric space.

j=1

(x j y j ) 2

1/2

.

Proof.

Since 0 (x j y j ) 2 2(x

2

j

+ y

2

j

) we have that

0

j=1

(x j y j ) 2 2

j=1

x

2

j

+ 2

j=1

y

2

j

< .

by

This proves (M1). (M2) and (M3) follow immediately from Deﬁnition 1.1. To prove (M4) we let z 2 . Let n N be arbitrary and let a i = x i z i , b i = z i y i . Since

n

i=1

(a i + b i ) 2 1/2

n

i=1

a

2

i

1/2 +

n

i=1

b

2

i

1/2

()

we have that

n

i=1

(x i y i ) 2 1/2 d(x, z) + d(z, y).

We obtain (M4) by letting n → ∞. To prove () we note that this equivalent to

n

i=1

a i b i

n

i=1

1/2 a i 2

n

i=1

b

2

i

1/2

.

The above follows directly from the fact that for λ R

n

i=1

(λa i + b i ) 2 = 2 + 2+ B 0,

where A = n

i=1 a i

2

, B = n

i=1 b i

2

have that H 2 AB.

, H

= n

i=1

a i b i . Since the quadratic in λ is non-negative we

Theorem 1.7 Let X be a non-empty set, and let B (X ) be the set of all bounded real-valued functions from X to R. Deﬁne d : B(X) 2 R by

d(f, g) = sup

xX |f(x) g(x)|.

Then d is a (supremum) metric on B(X).

 Proof. Since f, g ∈ B(X) there exist constants c 1 , c 2 such that |f (x)| ≤ c 1 , |g(x)| ≤ c 2 on X. Then 0 ≤ |f (x) − g(x)| ≤ |f (x)| + |g(x)| ≤ c 1 + c 2 . Hence

0 sup

xX |f(x) g(x)| ≤ c 1 + c 2 .

This proves (M1). (M2) and (M3) follow immediately from Deﬁnition 1.1. To prove (M4) we let h B(X). Then there exist c 3 such that |h(x)| ≤ c 3 on X. Hence

|f(x) g(x)|

and

≤ |f(x) h(x)| + |h(x) g(x)| ≤ sup xX |f (x) h(x)| + sup xX |h(x) g(x)| = d(f, h) + d(h, g),

xX |f (x) g(x)| ≤ d(f, h) + d(h, g).

sup

Deﬁnition 1.8 Let (X, d) be a metric space and let Y be a non-empty subset of X. The restriction of d : X × X R to d : Y × Y R induces a metric on Y .

Deﬁnition 1.9 (a) Let X be a vector space over R (or C). A mapping

 · : X → R is called a norm if (N1) (∀x ∈ X) [0 ≤ x < ∞] . (N2) (∀x ∈ X) [ x = 0 ⇔ x = 0] . (N3) (∀x ∈ X, ∀λ ∈ R(or C) ) [ λx = |λ| x ] . 2

(N4)

(x X, y X) [ x + y x + y ] .

(b) X supplied with a norm is called a normed space.

Proposition 1.10 Let (X, · ) be a normed space. Then X is a metric space with the metric deﬁned by

x, y X d(x, y) = x y .

Proposition 1.11

Proof.

Hence

By (M4)

x, y, z X

:

|d(x, z) d(y, z)| ≤ d(x, y).

d(x, z) d(x, y) + d(y, z).

d(x, z) d(y, z) d(x, y).

By reversing the role of x and y, and by (M3)

d(y, z) d(x, z) d(y, x) = d(x, y).

 ✷ Proposition 1.12 ∀x, y, u, v ∈ X : |d(x, u) − d(y, v)| ≤ d(x, y) + d(u, v). Proof. d(x, u) ≤ d(x, y) + d(y, u) ≤ d(x, y) + d(y, v) + d(v, u) = d(x, y) + d(y, v) + d(u, v) ⇒ d(x, u) − d(y, v) ≤ d(x, y) + d(u, v). By reversing the roles of x and y and of u and v respectively, d(y, v) − d(x, u) ≤ d(y, x) + d(v, u) = d(x, y) + d(u, v). ✷

Example 1.13 Let (X 1 , d 1 ), (X 2 , d 2 ) be metric spaces. Let X = X 1 ×X 2 , and deﬁne d : X 2 R by

(x 1 , x 2 ), (y 1 , y 2 ) X

d((x 1 , x 2 ), (y 1 , y 2 )) = d 2 (x 1 , y 1 ) + d 2 2 (x 2 , y 2 ).

1

Then (X, d) is a metric space: the Cartesian product of (X 1 , d 1 ) and (X 2 , d 2 ).

3

2

Convergence in metric spaces.

Deﬁnition 2.1 Let (X, d) be a metric space. A sequence of points x 1 , x 2 , X is convergent if there is an element x X such that

We write

lim

n x n = x or x n x.

n d(x n , x) = 0.

lim

(shorthand (x n ))in

Example 2.2 Let (R N , d) be the standard Euclidean space. Then

x (n) = (x

(n)

1

,

.

.

. , x

(n)

N

) x = (x 1 ,

, x N ) if and only if x

(n) x j for all 1 j N .

j

Proof.

N

(x

j=1

1/2

N

j=1

d((x

Also for j = 1,

(n)

1

,

.

.

, N

. , x

(n)

N

), (x 1 ,

, x N )) =

|x

(n)

j

(n)

j

x j ) 2

|x

(n)

j

 x j |. ✷ : X → Y , be a

x j | ≤ d(x (n) , x).

Deﬁnition 2.3 Let X be a set and let (Y, d) be a metric space.

sequence of functions. Then (f n ) converges pointwise if there exists a function f : X Y such that

Let (f n ), f n

(x X) [ lim

n d(f n (x), f (x)) = 0].

(f n ) converges uniformly if there exists a function f : X Y such that

or

lim

n→∞

(ε > 0) (N N)

sup

xX d(f n (x), f (x)) = 0,

(n N ) =sup

xX d(f n (x), f (x)) < ε .

Remark 2.4 Uniform convergence implies pointwise convergence. The converse need not be true. Convergence of a sequence of (f n ), f n B(X), n = 1, 2, · · · with the supremum metric is nothing but uniform convergence.

Example 2.5

1. However, (f n ) does not converge uniformly since

Let f n : (0, 1) R, f n (x) = x n . Then f n f

x(0,1) |f n (x) f (x)| =

sup

x(0,1) |f n (x)| =

sup

pointwise, where f (x) = 0, 0 < x <

sup

x(0,1)

x n = 1.

Let f n : [0, ) R,

f n (x) =

n+x . Then f n f pointwise, where f (x) = x, x 0. However,

nx

(f n ) does not converge uniformly since

f n n

2

f n

2

= n 6 → ∞.

Let f n : [0, ) R,

f n (x) =

converges uniformly to f since

nx 2 x .

1+n

Then f n f

pointwise, where f (x) = 0, x 0. (f n )

x[0,) |f n (x) f(x)| ≤

sup

1

n .

4

Let f n : [0, 1] R be given by

f n (x) =

1

n 2 + x 2 .

Then f n 0 in B[0, 1] with the supremum metric.

x[0,1] |f n (x) 0| =

sup

sup

x[0,1]

Let g n : [0, 1] R be given by

g n (x) =

1 1

n 2 + x 2 n 2

2nx

1 + n 2 x 2 .

0 (n → ∞)

Then g n does not converge in B[0, 1] with the supremum metric. g n (x) 0 for every x [0, 1]. But:

x[0,1] |g n (x) 0| = 1

sup

= 0.

Theorem 2.6 Let (X, ρ), (Y, d) be metric spaces, and let f n : X Y, f n f converge uniformly on X. If c is a point at which each f n is continuous, then f is continuous at c.

Proof. Let N N be such that n N implies sup xX d(f n (x), f (x)) <

ρ(c, x) < δ implies d(f N (x), f N (c)) <

3 ε . Then for ρ(c, x) < δ

3 ε . Let δ be such that

d(f (c), f (x)) d(f (c), f N (c)) + d(f N (c), f N (x)) + d(f N (x), f (x)) < ε

Hence f is continuous at c.

Deﬁnition 2.7 Let f n :

R, if the partial sums s n = f 1 + ··· + f n converge uniformly to s.

X R, n = 1, 2,

Then n=1 f n converges uniformly to s, s : X

Theorem 2.8 If the series n=1 f n of real-valued functions converges uniformly to s, and if each f n is continuous at c then s is continuous at c.

Proof.

s n = f 1 + ··· + f n is continuous at c. The result follows by Theorem 2.6

Example 2.9 n=1 x n converges pointwise on (1, 1) to s(x) =

Hence sup x(1,1) |s n (x) s(x)| = sup x(1,1) | x n+1 | ≥ gence is not uniform.

1x . Note that s n (x) = 1x n+1

1

.

2 , and the conver-

1x

1

1

2 sup x(1,1) |x| n+1 =

1x

Example 2.10 Let f n : [0.1] R, f n (x) = nxe nx 2 .

function which is 0 on [0, 1].

Then (f n ) converges pointwise to the

Note that

1 n )| ≥

The convergence is not uniform since |f n (

as n → ∞, and so √ n .
e

0

f n (x)dx = 1

1

2 (1 e n )

2

1

0

f n (x)dx

1

0

1

0dx.

Theorem 2.11 Let f n : [a, b] R, n = 1, 2,

which converges uniformly to f on [a, b].

be a sequence of Riemann integrable functions

b f n (x)dx

Then f is Riemann integrable, and

a

a

f (x)dx.

b

Proof. For each ε > 0 there exists N N such that n N, x [a, b] implies |f n (x) f (x)| < ε. Hence for any partition P of [a, b], and for n N

U(P, f) U(P, f n ) + ε(b a), L(P, f) L(P, f n ) ε(b a).

5

Since f N is Riemann integrable, there exists a partition P of [a, b] such that U ( P, f N )L( P, f N ) < ε. Then

U( P, f) L( P, f)

< U( P, f N ) + (b a) L( P, f N ) + ε(b a) < ε(1 + 2(b a)).

Since ε > 0 was arbitrary, f is Riemann integrable. Hence for n N

a

b

f (x)dx

=

inf{U (P, f ) : P a partition of [a, b]}

inf{U (P, f n ) : P a partition of [a, b]} + ε(b a)

Similarly, for n N

=

a

b

a

b

f n (x)dx + ε(b a).

f (x)dx

a

b

f n (x)dx ε(b a).

Combining these two inequalities we have for n N

|

a

b

f (x)dx

a

b

f n (x)dx| ≤ 2ε(b a).

Remark 2.12 Let f n : [0, ) R, f n (x) =

f n (x)dx = 1.

function on [0, ). However,

0

n 1 e x/n .

Then (f n ) converges uniformly to the 0

Remark 2.13 If f n f pointwise on [a, b], and f n is Riemann integrable for all n N. Then f

} be an enumeration of the rationals

f n (x)dx =

in [0, 1], f n (x) =

need not to be Riemann integrable. For example, let {q 1 , q 2

1,

0,

x ∈ {q 1 , otherwise.

,

q n },

1

Then f n is Riemann integrable with

0

0, f n f pointwise with f (x) =

But U (P, f ) = 1, L(P, f ) = 0, and f is not Riemann integrable.

1,

0,

Q [0, 1]

x

x [0, 1] Q.

Corollary 2.14 If n=1 f n converges uniformly to s on [a, b], and if each f n is Riemann inte-

grable on [a, b]. Then s is integrable, and

a

s(x)dx =

b

n=1 a

b

f n (x)dx.

Proof.

= lim n n

Put s n = f 1 + ··· + f n . Then

a

b

k=1 a

f k (x)dx =

k=1 a

b

s(x)dx = lim n

a

f k (x)dx.

b

b

s n (x)dx

Theorem 2.15 Let (f n ) be a sequence of real-valued diﬀerentiable functions on [a, b]. Suppose that (f n (x)) converges for at least one x [a, b], and that f n converges uniformly on [a, b]. Then (i) f n converges uniformly on [a, b], (ii) f = lim n f n is diﬀerentiable on [a, b], and for all x [a, b], f (x) = lim n f n (x).

Proof. Let lim n f n = g, and let ε > 0. Since (f n ) converges uniformly on [a, b], there exists

N

m,

N such that sup axb |f n (x) g(x)| < ε for n N . Hence sup axb |f n (x) f m (x)| ≤ 2ε for

n N . If ξ 1 , ξ 2 are points of [a, b] then, by the mean value theorem for f n f m

f n (ξ 1 ) f m (ξ 1 ) (f n (ξ 2 ) f m (ξ 2 )) = (ξ 1 ξ 2 ).{f n (ξ) f m (ξ)}

where ξ [ξ 1 , ξ 2 ]. So for all m, n N ,

(1)

|f n (ξ 1 ) f m (ξ 1 ) (f n (ξ 2 ) f m (ξ 2 ))| ≤ 2|ξ 1 ξ 2 |ε.

6

If x 0 is a point for which (f n (x 0 )) converges, then there is an N N such that for m, n N

(2)

|f n (x 0 ) f m (x 0 )| ≤ ε.

Let x be any point of [a, b]. By (1), (2) with ξ 1 = x, ξ 2 = x 0 we have for m, n max{N, N },

(3)

|f n (x) f m (x)| ≤ (1 + 2|x x 0 |)ε (1 + 2(b a))ε.

Hence lim m f m (x) exists for all x [a, b]. Taking the limit m → ∞ in (3) we see that (f n )

converges uniformly on [a, b]. Let a < c < b, and let h

By (1) with ξ 1 = c + h, ξ 2 = c we have for m, n N

= 0 be such that c + h [a, b].

f n (c + h) f n (c)

f m (c + h) f m (c)

h

h

2ε.

Since f m (c + h) f m (c) f (c + h) f (c) as m → ∞, the above inequality shows that for n N

f n (c + h) f n (c)

f(c + h) f(c)

h

h

2ε.

By the deﬁnition of the derivative f N (c) there exists on δ > 0 such that

(0 < |h| ≤ δ) =

f N (c + h) f N (c)

h

f N (c) ε .

Then for 0 < |h| ≤ δ

f(c + h) f(c)

h

g(c)

f N (c)

f N (c + h) f N (c)

+

h

Hence f (c) exists and f (c) = lim n f n (c).

+

f(c + h) f(c)

f N (c + h) f N (c)

h

h

|f N (c) g(c)| ≤ 2ε + ε + ε.

Corollary 2.16 Let u n : [a, b] R be such that

(i) (x 0 [a, b])

n=1

u n (x 0 ) is convergent .

(ii) (n N) u n is diﬀerentiable.

(iii) The series

n=1

u n (x) is uniformly convergent on [a, b].

Then the series

n=1

u n (x)

converges uniformly to s(x) and

s (x) =

n=1

u n (x).

 Theorem 2.17 If x n → x in (X, d) and x n → y in (X, d) then x = y. Proof.

0 d(x, y) d(x, x n ) + d(x n , y) 0

7

=x = y.

Theorem 2.18 Let x n x and y n y in (X, d). Then d(x n , y n ) d(x, y).

Proof.

0 ≤ |d(x n , y n ) d(x, y)| ≤ d(x, x n ) + d(y n , y) 0.

Deﬁnition 2.19 Let (X, d) and (X, e) be metric spaces. The metrics d and e are equivalent if

d(x n , x) 0 ⇐⇒ e(x n , x) 0.

Remark 2.20 If there are c 1 , c 2 > 0 such that

x, y X

then d and e are equivalent.

c 1 d(x, y) e(x, y) c 2 d(x, y)

Example 2.21 On R 2

and

are equivalent.

d((x 1 , y 1 ), (x 2 , y 2 )) = (x 1 x 2 ) 2 + (y 1 y 2 ) 2

e((x 1 , y 1 ),

(x 2 , y 2 )) = |x 1 x 2 | + |y 1 y 2 |

Proof.

1

2 e((x 1 , y 1 ), (x 2 , y 2 ))

d((x 1 , y 1 ), (x 2 , y 2 )) e((x 1 , y 1 ), (x 2 , y 2 )).

3 Limit points, open sets, closed sets

Let (X, d) be a metric space, x 0 X, r > 0.

Deﬁnition 3.1 The set B(x 0 ; r) = {x X

:

d(x, x 0 ) < r} is the open ball with center x 0 and

B(x 0 ; r) = {x X

: d(x, x 0 ) r} is the closed ball with center x 0 and radius r.

Example 3.2 Let R 1 have its usual metric, a R 1 . Then B(a; r) = (a r, a + r),

B(a; r) = [a r, a + r]. Similarly for R n , a R n ,

r 2 }, B(a; r) = {(