100% found this document useful (1 vote)
512 views17 pages

Chapter 1 Metric Spaces

The chapter introduces metric spaces. A metric space is a pair (X, d) where X is a set and d is a metric or distance function on X that satisfies four properties. Several examples of metric spaces are provided including the real line, Euclidean plane, sequence spaces, function spaces, and discrete metric spaces. Exercises are given to show that other combinations of sets and metrics form metric spaces. The chapter continues with further examples of metric spaces including sequence spaces and spaces of bounded functions. Auxiliary results like inequalities are also proved.

Uploaded by

ribeiro_sucesso
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
100% found this document useful (1 vote)
512 views17 pages

Chapter 1 Metric Spaces

The chapter introduces metric spaces. A metric space is a pair (X, d) where X is a set and d is a metric or distance function on X that satisfies four properties. Several examples of metric spaces are provided including the real line, Euclidean plane, sequence spaces, function spaces, and discrete metric spaces. Exercises are given to show that other combinations of sets and metrics form metric spaces. The chapter continues with further examples of metric spaces including sequence spaces and spaces of bounded functions. Auxiliary results like inequalities are also proved.

Uploaded by

ribeiro_sucesso
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 17

Chapter 1 Metric Spaces

1.1 Metric Space

1.1-1 Definition. A metric space is a pair ( X, d ), where X is a set and d is a


metric on X; that is a function on XX such that for all x, y, z X, we
have:
(M1) d( x, y )  0.
(M2) d( x, y ) = 0 if and only if x = y.
(M3) d( x, y ) = d( y, x ).
(M4) d( x, y )  d( x, z ) + d( z, y ).

Notes: 1) For any nN and any x1, x2, ……., xn X, we have
d(x1, xn )  d(x1, x2 ) + d(x2, x3 ) + ………. + d(xn-1, xn ).
~ ~
2) A subspace (Y, d ) of ( X, d ) is a subset Y of X with d = dY Y ( the
~
restriction of d on YY ), and d is called the metric induced on Y
by d.

Examples
1.1-2 Real line R. ( R, d ) is a metric space; where R is the set of real numbers
and for any x, y  R d( x, y ) = | x – y |.
Proof. Left to the reader
1.1-3 Euclidean plane R2. (R2, d1 ) is a metric space; where for any x = (1 , 2 )
and y = ( 1 ,  2 ) in R2 d1( x, y ) = | 1   1 |  |2   2 | .
Proof. Left to the reader
1.1-4 Sequence space  . Consider the set  of all bounded sequences of
complex numbers. Then (  , d ) is a metric space, where for any x = (i )
and y = ( i ) in  d( x, y ) = sup | i   i | .
iN

Proof. It is easy to check that (M1)-(M3) are satisfied. To check (M4),


note that for every x = (i ) , y = ( i ) and z = ( i ) in  , and for every
iN, we have | i   i |  | i  i | + | i   i |  d( x, z ) + d( z, y ). Take
the supremum of the left hand side over N to get the result
1.1-5 Function space C[a, b]. Let C[a, b] be the set of all real valued
functions that are continuous on J = [a, b]. For any x and y in C[a, b]
define d( x, y ) = max | x(t )  y(t ) | . Then ( C[a, b] , d ) is a metric space.
tJ

Proof. It is easy to check that (M1)-(M3) are satisfied. To check (M4),


note that for every x, y, z  C[a, b], and for every tJ, we have
| x(t) – y(t) |  | x(t) – z(t) | + | z(t) – y(t) |  d( x, z ) + d( z, y ). Take
the maximum of the left hand side over J to get the result

1
1.1-6 Discrete metric space. Suppose that X is a non-empty set. For every x,
 0 if x=y
y  X, define d(x, y) =  . Then ( X, d ) is a metric space.
1 if x  y
Proof. It is easy to check that (M1)-(M3) are satisfied. To check (M4),
let x, y, z  X be arbitrary, then we have two cases:
case 1. x = y, it is trivial that d( x, y )  d( x, z ) + d( z, y ),
case 2. x  y, then d(x, y) = 1. But then either ( z  x and z = y ), ( z  x
and z  y ) or ( z = x and z  y ), we have d(x, z) + d(z, y)  1. This
completes the proof
Exersise 1.1
1. Show that ( Rn, d1 ) and ( Rn, d2 ) are metric spaces; where for any x =
(1 , 2 ,......, n ) and y = ( 1 ,  2 ,......,  n ) in R , d1( x, y ) = | 1 - 1 | + ……..
n

+ | n - n | and d2( x, y ) = max { | 1 - 1 |, ..….. , | n - n | }. Then show that


d2(x, y)  d1(x, y)  n d2(x, y) .
2. . Suppose that that (X1, d1), . . . , (Xn, dn) are metric spaces and suppose that x
= (1 , 2 ,......, n ) and y = ( 1 ,  2 ,......,  n ) be any elements in X1 × ..….. ×Xn.
Show that ρ1, and ρ2 are metrics on the set X1 × ..…. . × Xn, and ρ2(x, y) 
ρ1(x, y)  n ρ2(x, y), where ρ1(x, y) = d1( 1 , 1 ) + …….. + dn ( n , n ),
and ρ2( x, y ) = max { d1( 1 , 1 ), ..….. , dn ( n , n )}.

H. W. 3-5, 7, 8, 10-15. H.W.* 4, 5, 8, 10 + 1, 2 above

1.2 Further Examples of Metric Spaces

1.2-1 Sequence Space s. Let s be the set of all sequences of complex numbers.

Define d( x, y ) =  i 1
|i  i |
2 (1|i  i |)
i
, for any x = (i ) and y = ( i ) in s.
Show that ( s, d ) is a metric space.
Proof. It is easy to check that (M1)-(M3) are satisfied. To check (M4), we use
t
the auxiliary real valued function f defined on R\{-1} by f(t) = . It is easy
1 t
1
to see that f is an increasing function where its derivative f /(t) = is
(1  t ) 2
positive. Now, for every x = (i ) , y = ( i ) and z = ( i ) in s, and for every iN,
we have | i   i |  | i  i | + | i   i | . Then f( | i   i | )  f( | i  i | +
| i   i | | i   i |  |  i   i | | i   i |
| i   i | ). That is,  =
1 | i   i | 1 | i  i |  | i   i | 1 | i  i |  | i   i |
| i   i | | i   i | | i   i |
+  + . Finally multiply both
1 | i  i |  | i   i | 1 | i   i | 1 | i   i |
, then take the sum over i from 1 to ∞ to get d( x, y )  d( x, z ) +
1
sides by 2i
d( z, y ). This completes the proof

2
1.2-2 Space B(A) of bounded functions. Let B(A) be the set of all bounded
functions defined on a set A. Define d( x, y ) = sup | x(t )  y(t ) | , for any x
tA

and y in B(A). Show that (B(A), d ) is a metric space.


Proof. See the proof of example 1.1-4

1 1
1.2-2 Auxiliary inequality. Let p > 1 and define q by p + q = 1. If  and  are
p q
nonnegative numbers, then    p + q . [ p and q are then called conjugate
exponents].
Proof. The inequality is trivial in the case that  = 0 or  = 0. Now let  and 
1 1 pq
be positive. Since p + q = 1, then we have pq = 1, pq = p + q, and
pq – p – q + 1= 1. Hence p( q -1) – ( q – 1) = 1 and ( p – 1 )( q -1) = 1. So that
for real numbers t and u we have that u = t p-1 implies that t = u 1/(p-1) = u (q-1).
Since   is the area of the rectangle in Fig. 1 below [ see th text book ], then
 

t dt   u q 1du =
p 1 p q
by integration we have    p + q 
0 0
p
Notation. well denote the set of all sequences of numbers (i ) such that

| 
i 1
i | p is finite, where p  1.

p
1.2-3 Holder Inequality. Let p and q be as in 1.2-2 above. If (i )  and
  
, then  | i i |  ( |  k | ) ( |  m | ) .
1 1
q p p q q
( i ) 
i 1 k 1 m 1
Proof. The inequality is trivial in the case that (i ) or ( i ) is zero. Now let (i )
~
i
and  i 
~
i
and ( i ) be nonzero. For iN, set i    1 .
| k |q )
1
| k | p )
q
( p (
k 1 k 1
~ ~
~ ~ | i | p | i |q
By auxiliary inequality, | i  i |  p + q . Take the sum over i from
 ~ ~ 1  ~ p 1  ~ q
1 to ∞ of both sides to get,  | i  i | 
i 1
( |  i | ) +
p i 1
( |  i | ) =
q i 1
  
= 1. Hence  | i i |  ( |  k | ) ( |  m | ) 
1 1
1 1 p p q q
p + q
i 1 k 1 m 1

Note. If p = q = 2, we have the Cauchy Schwarz inequality


  

|   |  ( |  k | ) ( |  m |2 ) 2 .
1 1
2 2
i i
i 1 k 1 m 1

3
p
1.2-4 Minkowski Inequality. If p  1 and (i ) , ( i )  , then
  
( | i   i | p )  ( |  k | ) + (  |  m | ) .
1 1 1
p p p p p

i 1 k 1 m 1
Proof. For p =1 the inequality follows directly from the triangle inequality of
numbers. Let p > 1. For any iN,
| i   i | p = | i   i | | i   i | p 1  ( | i | + |  i | ) | i   i | p 1 . Take
n
the sum over i from 1 to any fixed nN of both sides to get | 
i 1
i   i |p 
n n

| 
i 1
i || i   i | p 1 + | 
i 1
i || i   i | p 1 ………………….. (1)

By Holder inequality
n n n

 | i || i   i | p 1  ( |  k | p ) p ( (| m   m | p 1 )q ) q
1 1

i 1 k 1 m 1
n n
= ( |  k | ) ( |  m   m |( p 1) q )
1 1
p p q

k 1 m 1
n n
= ( |  k | ) ( |  m   m | ) .
1 1
p p p q

k 1 m 1
n n n

|  || i   i |  ( |  k | ) ( |  m   m | ) .
1 1
p 1 p q p p
Similarly, i
i 1 k 1 m 1
Substitute the last two inequalities in (1) to get,
n n n

 | i   i | p  ( |  k | p ) p ( |  m   m | p ) q +
1 1

i 1 k 1 m 1
n n n n
( |  k | ) ( |  m   m | ) = { ( |  k | ) + ( |  k | p ) p }
1 1 1 1
p p p q p p

k 1 m 1 k 1 k 1
n n n
( |  m   m | p ) q . Hence ( | i   i | p )  ( |  k | ) +
1
1 1
q p p
1

m 1 i 1 k 1
n
( |  k | p ) p . By noting that 1 -
1
1
q = p, and letting n  ∞ you get
k 1
  
( | i   i | p )  ( |  k | ) + (  |  m | ) 
1 1 1
p p p p p

i 1 k 1 m 1

p p
1.2-5 Sequence space . Consider the set of all sequences of numbers

(i ) such that | 
i 1
i | p is finite, where p  1. Then ( p
, d ) is a metric

4
p
space, where for any x = (i ) and y = ( i ) in

d( x, y ) = ( | i   i | )
1
p p
.
i 1
Proof. It is easy to check that (M1)-(M3) are satisfied. To check (M4), use
p
Minkowski inequality, where for every x = (i ) , y = ( i ) and z = ( i ) in ,
 
d( x, y ) = ( | i   i | ) ( [| i -i | + | i - i |] p )
1 1
p p p

i 1 i 1
 
 ( | i   i | ) + ( |  i   i | )
1 1
p p p p
= d( x, z ) + d( z, y )
i 1 i 1

Note. If p = 2, we have an important space called the Hilbert sequence space


2
.

In the following examples you can use Minkowski inequality to prove the
triangle inequality.
1.2-6 Euclidean plane R2. (R2, d ) is a metric space; where for any
x = (1 , 2 ) and y = ( 1 ,  2 ) in R2, d( x, y ) = (1   1 )2  (2   2 )2 .
This metric is called the Euclidean metric.
Proof. Left to the reader
1.2-7 Three dimensional Euclidean space R3. (R3, d ) is a metric space;
where for any x = (1 , 2 , 3 ) and y = ( 1 ,  2 ,  3 ) in R3 d( x, y ) =
(1   1 )2  (2   2 )2  (3   3 )2 . This metric is called the Euclidean metric.
Proof. Left to the reader
1.2-8 Euclidean space Rn, unitary space Cn, complex plane C.
a) ( Rn, d ) is a metric space; where for any x = (1 , 2 ,......, n ) and
y = ( 1 ,  2 ,......,  n ) in Rn, d( x, y ) = (1   1 )2  (2   2 )2  .........  (n   n )2 .
This metric is called the Euclidean metric.
b) Also ( Cn, d ) is a metric space, where for any x = (1 , 2 ,......, n )
and y = ( 1 ,  2 ,......,  n ) in Cn d( x, y ) = | 1   1 |2  | 2   2 |2 ......... | n   n |2 .
When n = 1 this is the complex plane C.
Proof. Left to the reader

Exersise 1.2
1. Let ( Rn, d1 ) be as in problem 1 Exersise 1.1. Show that d(x, y)  d1(x, y)
 n d(x, y), where d is the Euclidean metric on Rn.
2. In the notation of problem 2 in Exersise 1.1, show that ρ(x, y)  ρ1(x, y) 
n ρ(x, y), where ρ(x, y) = d12 (1 ,  1 )  .........  d n 2 (n ,  n ) .

H. W. 1-3, 6-12. H.W.* 8, 10 + 1, 2 above.

5
1.3 Open Sets, Closed Sets, Neighborhood

1.3-1 Definition. Let ( X, d ) be a metric space, x0 X and r be a positive real


number. Define the sets
a. B(x0; r) = { xX : d(x, x0) < r } as an open ball.
~
b. B (x0; r) = { xX : d(x, x0)  r } as a closed ball.
c. S(x0; r) = { xX : d(x, x0) = r } as a sphere.
Where x0 is called the center and r is the radius.
~
From the definition one can easily see that, S(x0; r) = B (x0; r) - B(x0; r).
From now on, we can let X to denote the metric space ( X, d ).
1.3-2 Definition. A subset M of a metric space X is said to be open if it
contains a ball about each of its points. A subset K of X is said to be
closed if its complement is open.
It is easy to see from the definition above that an open ball is an open set and a
closed ball is a closed set.
1.3-3 Definition. Let X be a metric space. An open ball B(x0; ) of radius  is
called an -neighborhood of x0 X . A neighborhood of x0, ( denpted by
N(x0) ) is a subset of X which contains an -neighborhood of x0.
1.3-4 Definition. Let X be a metric space. x0 is called an interior point of a
set M  X if M is a neighborhood of x0. The set of all interior points of
M is denoted by Int(M) or M 0.
It is easy to see that Int(M) is an open set.
1.3-5 Theorem. Let X be a metric space and  be the collection of all open
subsets of X. Then the following properties hold:
(T1) φ, X.
(T2) The union of any members of  is a gain in .
(T3) The intersection of finitely many members of  is a gain in .
Proof. (T1) is obvious. To prove (T2), let U be the union of arbitrary elements
( open subsets of X ) in  and let x be any element in U. Then x belongs to at
least one of these open sets, call this set M. Hence there is a ball B  M and
xB. Thus M  U, so that U is open; that is U. Finally, let y be any element
in the intersection of M1, M2, ………., Mn, then each Mi contains a ball about
y and the smallest of these balls is contained in the intersection. Therefore, the
intersection is open; that is it belongs to .
1.3-6 Definition. A topological space ( X,  ) is a set X and a collection  of
subsets such that  satisfies (T1), (T2), and (T3).
From Theorem 1.3-5 and the above definition we can say that “ a metric space
is a topological space.”
~
1.3-7 Definition. Let ( X, d ) and (Y, d ) be metric spaces. A mapping
T : X Y is said to be continuous at x0X if for every  > 0 there is
~
 > 0 such that d ( Tx, Tx0 ) <  for all x satisfying d( x, x0 ) < . T is
said to be continuous if it is continuous at every point of X.

6
~
1.3-8 Theorem. Let ( X, d ) and (Y, d ) be metric spaces. A mapping T: XY
is continuous if and only if the inverse image of any open subset of Y is
an open subset of X.
Proof. Suppose that T is continuous. Let S be any open set in Y and let G =
T -1(S). If G = , then it is open. Let G   and let x0 be any element in G.
Then there is y0S such that y0 = Tx0. However; S is an open set, then there is
a ball B(y0; )  S. Since T is continuous at x0 then there is  > 0 such that for
~
all x, d( x, x0 ) <  implies d ( Tx, Tx0 ) < . Now let xB(x0; ) then d( x, x0 )
~
<  and so d ( Tx, Tx0 ) < ; that is TxB(y0; ). However; B(y0; )  S, then
TxS and so xG = T -1(S). Hence B(x0; )  G. Therefore, G = T -1(S) is an
open subset of X.
Conversely, assume that the inverse image of every open set in Y is an open set
in X. Let x0 be any fixed point in X and let  > 0 be arbitrary. Since B(Tx0; ) is
an open set in Y, then T -1(B(Tx0; )) is an open set in X, then there is > 0 such
that B(x0; )  T -1(B(Tx0; )) . Therefore, for every  > 0 there is > 0 such
that for all x if d( x, x0 ) < , then xB(x0; ) and x T -1(B(Tx0; )). Hence
~
Tx B(Tx0; ) and so d ( Tx, Tx0 ) < . Therefore, T is continuous at x0. Since
x0 was arbitrary in X then T is continuous on X

1.3-9 Definition. Let M be a subset of metric space X. x0X is called an


accumulation point of M if for every neighborhood N(x0) of x0,
___
N(x0) \ {x0}  M  φ. The closure of M ( denoted by M ) is the set
consisting of M with its accumulation points.
_________ ~
Note. B(x 0 ; r) need not equal to B (x0; r). To see this consider the discrete
metric space (X, d) with X has at least two points. It is easy to see that
____________ ~
B(x 0 ; 1) = { x0 }  X = B (x0; 1).

1.3-10 Definition. A subset M of a metric space X is said to be dense in X if


___
M = X. X is said to be separable if it has a countable subset which is
dense in X.
Hence, M is dense in X if an only if every ball in X will contains points of M.

Examples
1.3-11 Real line R. The real line is separable.
Proof. Left to the reader

1.3-12 Complex Plane C. The complex plane is separable.


Proof. Consider the countable set M = { r1 + ir2 : r1, r2 are rational numbers}
___
and show that M = C

7
1.3-13 Discrete metric space. A discrete metric space X is separable if and
only if X is countable.
Proof. Suppose that X is separable, then there is a countable subset M of X
___
such that M = X. However, any subset of a discrete metric space is closed
hence M = X. Therefore, X is countable. Conversly, suppose that X is
___
countable then X = X. Hence X is separable

p p
1.3-14 The space . The space with 1  p <  is separable.
p
Proof. Consider the set M = { ( r1, r2,….., rn, 0,0,0,……)  : ri’s are rational
___
p
numbers and n N}, which is a countable set (how?). We show that M = .
p
Let x = (i )  be arbitrary. Then for every  > 0 there is n = n()N such

 p
that 
j  n 1
|  j |p <
2
. Since the rationals are dense in R, then for each  i

there is a rational number rj close to  i . Hence we can find y M satisfying



 p n 

 |  j  rj | p <
j 1 2
. Therefore, [ d(x, y) ] p =  |  j  rj | p +
j 1
 |
j  n 1
j |p <

 p
 p ___
+ =  p . Hence d(x, y) <  and so yB(x; ). Then x M .
2 2
___
p p
Therefore, M = . This completes the proof of the separability of 

Note. For any y  [0, 1] there is a sequence (i ) such that  i {0, 1} for all i
and y = 1 +  22 +  33 + …… is a binary representation of y. Moreover
2 2 2
different y  [0, 1] have different binary representation. [ See Introduction to
real analysis, R. G. Bartle & D. R. Sherbert, pp56-59].

 
1.3-15 The space . The space is not separable.
 
Proof. Let M be any dense subset of , and let S = { (i )  :  i {0, 1} }.
Since [0, 1] is uncountable and for each y  [0, 1] there corresponds (i )  S and
different y’s in [0, 1] have different sequences in S [ see the note above ]. Hence
S is uncountable. For any x = (i ) , y = ( i ) in S, x  y we have B(x; 1 ) 
3
B(y; 1
) = . Hence we have uncountable number of balls that are not intersect.
3

However, M is dense in then each of the above balls must contain an

element of M. Thus M must be uncountable. Since M was arbitrary, then is
not separable

H. W. 1, 2, 4-6, 13-15. H. W.* 6, 14.

8
1.4 Convergence, Cauchy Sequences, Completeness

1.4-1 Definition. A sequence (xn) in a metric space ( X, d ) is said to be


convergent if there is xX such that lim d( xn, x ) = 0. In this case x is
n

called the limit of xn or (xn) converges to x and we write lim xn = x or


n

simply, xn  x. If (xn) is not convergent, it is said to be divergent.


Notes: 1) lim xn = x if and only if for every  > 0 there is k = k()N such that
n

xn B(x; ) for all n > k.


2) The limit of a convergent sequence must be a point of the space X in 1.4-1,
otherwise the sequence is divergent. To see this, let X = (0, 1) with the metric
d(x, y) = | x - y | and (xn) = ( 1 ). It is clear that ( 1 ) is divergent, because ( 1 )
n n n
wants to converge to 0 which is not in (0, 1).
1.4-2 Definition. A nonempty subset M of a metric space X is said to be
bounded if the diameter of M, (M) = sup{ d(x, y): x,yM } is finite.
The sequence (xn) in X is called a bounded sequence if the corresponding
point set is a bounded subset of X.
1.4-3 Lemma. Let ( X, d ) be a metric space. Then:
a) A convergent sequence in X is bounded and its limit is unique.
b) If xn  x and yn  y in X, then d(xn, yn)  d(x, y).
Proof. a) Suppose that xn  x . Then for  = 1, there is kN such that d(xn, x)
< 1 for all n > k. Let a = max { d(x1, x), d(x2, x), ……, d(xk-1, x) }. Then for all
n N we have, d(xn, x) < 1 + a. Hence for all n and m, d(xn, xm)  d(xn, x) +
d(xm, x) < 2 + 2a. Thus ( {xn})  2 + 2a , that implies (xn) is bounded.
Assume that xn  x and xn  z. Then 0  d(x, z)  d(x, xn) + d(xn, z)  0 as
n  Therefore, x = z, so that the limit is unique
b) Suppose that xn  x and yn  y. Then d(xn, yn)  d(xn, x) +d(x, y) +d(y, yn).
Hence d(xn, yn) - d(x, y)  d(xn, x) +d(y, yn). Similarly, d(x, y) - d(xn, yn) 
d(x, xn) +d(yn, y). Therefore, - [ d(xn, x) +d(y, yn) ]  d(xn, yn) - d(x, y) 
d(xn, x) + d(y, yn), that is, 0  | d(xn, yn) - d(x, y) |  d(xn, x) + d(y, yn)  0
as n . Hence d(xn, yn)  d(x, y) as n 
1.4-4 Definition. A sequence (xn) in a metric space ( X, d ) is said to be
Cauchy if for every  > 0 there is k = k()N such that d(xn, xm) <  for
all n, m > k. The space X is complete if every Cauchy sequence in X
converges.
1.4-5 Theorem. The real line and the complex plane are complete metric
spaces.
Note. For aR, R-{a}, and the set of all rational numbers Q are not complete.
1.4-6 Theorem. Every convergent sequence in a metric space is a Cauchy
sequence.
Proof. Suppose that xn  x . Then for every  > 0, there is k = k()N such
that d(xn, x) <  for all n > k. Hence for all n, m > k we have,
2

9
d(xn, xm)  d(xn, x) + d(x, xm) <  +  = . This shows (xn) is a Cauchy
2 2
sequence
1.4-7 Theorem. Let M be a nonempty subset of a metric space ( X, d). Then:
___
a) x M if and only if there is a sequence ( xn ) in M such that xnx.
b) M is closed if and only if the situation xnM, xnx implies that xM.
___
Proof. a) Let x M . If xM, then ( x, x, x, …..) is a sequence in M and
converges to x. If xM, it is an accumulation point of M. Hence for each n N,
B(x; 1 )  M   so there is xn B(x; 1 )  M. Hence d(xn, x) < 1 for all n.
n n n
Thus xnx as n .
___
Conversely, if ( xn ) in M such that xnx, if xM then x M . If xM, then for
every  > 0, there is k = k()N such that d(xn, x) <  for all n > k. Hence for
___
every  > 0, there is k = k()N such that B(x; )\{x}  M . Then x M 
___
b) By using a) above we have, M is closed if and only if M = M if and only if
the situation xnM, xnx implies that xM
1.4-8 Theorem. A subspace M of a complete metric space X is itself complete
if and only if the set M is closed in X.
___
Proof. Suppose that M is complete, and x M . Then by Theorem 1.4-7 (a),
there is a sequence ( xn ) in M such that xnx. Hence ( xn ) is Cauchy.
___
However, M is complete then xM. Therefore, M = M , that is M is closed.
Conversely, let M be a closed set and ( xn ) be Cauchy in M. By the
___
completeness of X, xnxX. Then x M , however, M is closed then xM.
Therefore, M is complete
1.4-9 Theorem. A mapping T : X  Y of a metric space ( X, d) into a metric
~
space (Y, d ) is continuous at x0X if and only if xnx0 implies
TxnTx0.
Proof. Suppose that T is continuous at x0X. Then for every  > 0, there is  > 0
~
such that d(x, x0) <  implies d (Tx, Tx0) < . Let xnx0. Then there is kN
~
such that d(xn, x0) <  for all n > k. Hence d (Txn, Tx0) <  for all n > k. Thus,
TxnTx0.
Conversely, Suppose that xnx0 implies TxnTx0. To prove that T is
continuous at x0 we use the contradiction. Suppose that T is not continuous at x0.
Then there is  > 0 such that for every  > 0 there is x  x0 with d(x, x0) < , but
~
d (Tx, Tx0)  . In particular, for  = 1 there is xnX satisfying d(xn, x0) < ,
n
~
but d (Txn, Tx0)  . Hence xnx0 but (Txn) does not converge to Tx0. This
contradicts the assumptiopn. Therefore, T is continuous at x0

H. W. 1, 2, 4-6, 8-10. H. W.* 6, 8.

11
1.5 Examples. Completeness Proofs

Examples

1.5-1 Completeness of R n and C n. Euclidean space R n and unitary space C n


are complete.
Proof. Let ( xm) be any Cauchy sequence in Rn, where xm = ( 1( m ) ,  2( m ) ,….,  n( m ) ).
Since ( xm) is Cauchy, then for every  > 0, there is k = k()  N such that
n
d(xm, xr) = ( (i
1
( m)
 i( r ) ) 2 ) 2
<  for all m, r > k ……………… (1)
i 1
n
Then for all m, r > k and all i = 1, 2,…,n (i( m)  i( r ) )2   (
i 1
i
( m)
 i( r ) )2 <  2.

So that |    | < . Hence for any fixed i, 1  i  n, the sequence ( i(1) , i(2) ,
i
( m)
i
(r )

……… ) is a Cauchy sequence of real numbers, then it converges say, to i ; that


is i( m) i as m . Then x = ( 1, 2 , …., n ) Rn. By (1), as r  we have
n
d(xm, x ) = ( (i  i ) )
1
  for all m > k. Hence xm xRn. Therefore,
( m) 2 2

i 1

R is complete. Similarly for C n


n

 
1.5-2 Completeness of . The space is complete.

Proof. Let ( xm) be any Cauchy sequence in , where xm = ( 1( m ) ,  2( m ) , …….).
Then for every  > 0, there is k = k()N such that d(xm, xn) = sup | i( m)  i( n ) | < 
iN

for all n, m > k. Then for all n, m > k and any fixed i N, |   i( n) | <  .… (2) i
( m)

Hence for any fixed i, the sequence ( i(1) , i(2) , ……… ) is a Cauchy sequence
of numbers ( real or complex ). Then it converges ( because both R and C are
complete ) say, to i that is i( m ) i as m . Using these infinitely limits

we define x = ( 1, 2 , …. ). We show that x and xm x as m . From
(2), as n  we have that for all m > k. | i  i |   ……………....…. (3)
( m)


Since xm = ( 1( m ) ,  2( m ) ,…..) then there is km such that | i( m ) |  km for all i.
Hence for any i, | i |  | i - i( m ) | + | i( m ) |   + km. Hence ( i ) is a

bounded sequence of numbers. Then x . By (3), we have d( xm, x ) =
sup | i  i |   for all m > k. This shows that xm x. But ( xm ) was
( m)

iN
 
arbitrary Cauchy sequence in . Hence is complete

11
1.5-3 Completeness of c. The space c consists of all convergent sequences of

complex numbers with the metric induced from the space is complete.
___
Proof. Let x = ( i ) c . Then there is a sequence ( xm) in c such that xm x,
where xm = ( 1( m ) ,  2( m ) , …….). Hence for every  > 0, there is kN such that

sup | i( m )  i | = d(xm, x) < 3 for all m  k. Then for all m  k and any fixed i
i N
 
N, | i( m )  i | < 3 . In particular for m = k and all i | i( k )  i | < 3 . Since xkc,
its terms i( k ) form a convergent sequence, such a sequence is Cauchy. Hence there

is n such that | i( k )   j( k ) | < 3 for all i, j > n. Therefore, for all i, j > n we have,
  
| i - j |  | i  i( k ) | + | i( k )   j( k ) | + |  j( k )   j | <
+ 3 + 3 = . 3
Hence x = ( i ) is a Cauchy sequence of numbers, and so by the completeness of C,
___
( i ) is convergent this shows that xc. Therefore, c = c . Hence c is a closed
 
subspace of . However, is complete, then by Theorem 1.4-8, c is complete

p p
1.5-4 Completeness of . The space is complete, where p is fixed and
1  p < .
p
Proof. Let ( xm) be any Cauchy sequence in , where xm = ( 1( m ) ,  2( m ) , …….).
Then for every  > 0, there is k N such that for all n, m > k we have,

d(xm, xn) = ( | i( m )  i( n ) | p )
1
p
<  …………………………………….……… (4)
i 1

It follows that for any fixed i N, | i( m)  i( n) | < . This shows that for any fixed i,
the sequence ( i(1) , i(2) , …… ) is a Cauchy sequence of numbers ( real or complex ).
Then it converges ( because both R and C are complete ) say, to i that is i( m ) i
as m . Using these infinitely limits we define x = ( 1, 2 , …. ). We show that
p
x and xm x as m . From (4), we have that for all m, n > k and all l,
l l

 | i(m)  i(n) | p <  p. Letting n , we obtain


i 1
| 
i 1
i
( m)
 i | p   p for all m > k

and all l. Now let l ; then for m > k | i 1
i
( m)
 i | p   p and

( | i( m )  i | p )
1
p
  ………………………………………………………….(5)
i 1

, then ( | i( m ) | p )
1
p
Since xm = (  i
( m)
) p
is finite. By Minkowski inequality and
i 1
  
 ( | 
1
+ ( | i( m ) | p )
1
by using (5), ( | i | )
1
 i | )
p p ( m) p p p p
i is finite. Hence x .
i 1 i 1 i 1

12
Furthermore, by (5) d( xm, x )   for all m > k. This shows that xm x. But
p p
( xm ) was arbitrary Cauchy sequence in . Hence is complete

1.5-5 Completeness of C[a, b]. The function space C[a, b] is complete.


Proof. Let ( xm) be any Cauchy sequence in C[a, b]. Then for every  > 0, there is
k N such that for all n, m > k we have, d(xm, xn) = max | xm (t )  xn (t ) | <  ….. (6),
tJ

where J = [a, b]. Hence for all n, m > k and any fixed t0 J, | xm (t0 )  xn (t0 ) | < .
This shows that ( x1(t0), x2(t0), …….. ) is a Cauchy sequence of real numbers,
then it converges ( because R is complete ) say, xm(t0) x(t0) as m . Then
for each t J we can associate a unique x(t)R. This defines a function x on J.
From (6) as n  max | xm (t )  x(t ) |   for all m > k; that means xm x as
tJ

m . Hence for all m > k and any tJ, | xm (t )  x(t ) |  , that means ( xm)
converges uniformly to x on J. However, xm’s are continuous on J. Hence x is
continuous on J; that is xC[a, b]. Therefore, C[a, b] is complete

Note. The above proof also proves the following theorem.

1.5-6 Theorem ( Uniform convergence ). Convergence xm x in the space


C[a, b] is a uniform convergence, that is ( xm) converges uniformly on [a, b]
to x.

Examples of incomplete metric spaces

1.5-7 Space Q. Let Q be the set of all rational numbers with the usual metric
d(x, y) = | x – y | for all x, y R. Then Q is not complete.
Proof. Let x be a fixed irrational number, then for any n N there is rnQ
1
such that x < rn < x + n . Hence rn  x as n . But xQ. Therefore, Q is
not complete

1.5-8 Space of Polynomials. Let X be the set of all polynomials over the real
interval J = [a, b] and define a metric d on X by d(x, y) = max | x(t )  y(t ) | .
tJ
Then (X, d) is not complete.
Proof. Left to the reader

13
1.5-9 The space of continuous functions. Let X be the set of all continuous
real-valued functions on J = [0, 1] and define a metric d on X by
1

d(x, y) =  | x(t )  y(t ) | dt . Then (X, d) is not complete.


0
Proof. Let ( xm) be a sequence of continuous functions on [0, 1], where
 0 if t  [0, 12 ]

xm(t) = m(t  12 ) if t  ( 12 , 12  m1 ) . For every  > 0, there is a natural
 1 if t  [ 12  m1 , 1]

1 1 1
number k   such that for all n > m > k we have, d(xm, xn) = m - n <
1
< , because d(xm, xn) is the area of the triangle in Fig.10 [ see the text
m
book].
1

Hence ( xm) is a Cauchy sequence. For any xX, d(xm, x) =  | xm (t )  x(t ) | dt =


0
2 m
1 1 1
2 1

 | x(t ) | dt +  | xm (t )  x(t ) | dt +  |1  x(t ) | dt . Since the integrals on the right


2 m
0 1 1 1
2
1
2

are nonnegative, then d(xm, x) 0 as m  would implies that  | x(t ) | dt = 0


0
1
1 2 1

and  |1  x(t ) | dt  0 as m . Hence  | x(t ) | dt =  |1  x(t ) | dt = 0.


2 m
1 1 0 1
2

However, x(t) is continuous, then x(t) = 0 on [0, ½) and x(t) = 1 on (½, 1].
But this is impossible for a continuous function. Therefore, d(xm, x) does not
converge to 0; that is (xm) does not converge to x. However x was arbitrary in
X, thus (xm) does not converge. Therefore, (X, d) is not complete

H. W. 1-5, 7, 8, 10-14. H. W.* 10, 13, 14 + Example 1.5-8.

14
1.6 Completion of Metric Spaces
~
1.6-1 Definition. Let ( X, d ) and (Y, d ) be metric spaces. Then
(a) A mapping T : X Y is said to be isometric or an isometry if T
~
preserves distances ; that is for all x, yX, d ( Tx, Ty ) = d( x, y ).
(b) The space X is said to be isometric with the space Y if there exists a
bijection isometry of X onto Y. The spaces X and Y are then called
isometric spaces.
1.6-2 Theorem ( Completion ). For a metric space ( X, d ) there exists a
 
complete metric space ( X , d ) which has a subspace W that is isometric
 
with X and is dense in X . This space X is unique except for isometries,
~ ~
that is, if X is any complete metric space having a dense subspace W
~ 
isometric with X, then X and X are isometric.
Proof. The proof is subdivided into the following four steps.
 
(a) Constructing ( X , d ).
___ 
(b) Constructing an isometry T of X onto W, where W = X .

(c) Proving the completeness of X .

(d) Proving the uniqueness of X , except for isometries.
 
(a) Constructing ( X , d ).
Let (xn) and (zn) be Cauchy sequences in X.
Define (xn) ~ (zn), if lim d(xn, zn) = 0. ( ~ is an equivalence relation, why?)…(1)
n
  
Let X be the set of all equivalence classes x , y , ….. of Cauchy sequences thus
  
obtained. Set d ( x , y ) = lim d( xn, yn )………………………………………(2),
n
    
where ( xn)  x and ( yn)  y and x , y  X . To prove that this limit exists we do
as follow: d( xn, yn )  d( xn, xm ) + d(xm, ym ) + d(ym, yn ). Then d( xn, yn ) -
d(xm, ym )  d( xn, xm ) + d(ym, yn ). Similarly, d(xm, ym ) - d( xn, yn )  d( xn, xm )
+ d(ym, yn ). Together, | d( xn, yn ) - d(xm, ym ) |  d( xn, xm ) + d(ym, yn ). Since
( xn) and ( yn) are Cauchy sequences, then for every  > 0, there is k such that for
 
all n, m > k we have, d( xn, xm ) < 2 and d(ym, yn ) < 2 , so that | d( xn, yn ) -
d(xm, ym ) | < . Hence ( d( xn, yn ) ) is a Cauchy sequence of real numbers, then
it converges. Thus lim d( xn, yn ) exists. Now, we show that this limit is
n
 
independent of the choice of ( xn)  x and ( yn)  y , in other words, if (xn) ~ (zn)

15
and (yn) ~ (wn), then by (1) | d( xn, yn ) - d(zn, wn ) |  d( xn, zn ) + d(yn, wn) 0

as n  which implies lim d( xn, yn ) = lim d(zn, wn ). We prove that d in (2) is
n n
    
a metric on X . It is clear that d satisfies (M1), (M3) and d ( x , x ) = 0.
  
Furthermore, if d ( x , y ) = 0, then lim d( xn, yn ) = 0 and so (xn) ~ (yn). Hence
n
 
the equivalence classes x and y are equal. Therefore, (M2) is satisfied. To prove
      
(M4), note that, for all ( xn)  x , ( yn)  y and ( zn)  z and x , y , z  X we have,
  
d( xn, yn )  d( xn, zn ) + d(zn, yn ) and, d ( x , y ) = lim d( xn, yn )  lim d( xn, zn )
n n
       
+ lim d(zn, yn ) = d ( x , z ) + d ( z , y ). Therefore, ( X , d ) is a metric space.
n
___ 
(b) Constructing an isometry T of X onto W, where W = X .
 
Define T:X W  X by T(b) = b , where W = T(X), and for each bX we associate
 
the class b  X which contains the constant Cauchy sequence (b, b, b, b,…....). For
   
any a, bX, d (Ta, Tb) = d ( a , b ) = lim d(a, b) = d(a, b). Hence T is an isometry
n
from X onto W. However, any isometry is 1-1, then X and W are isometric. We
___    
show that W = X . Let x  X and ( xn)  x . Then for every  > 0, there is k such that
  
d( xn, xk ) < 2 for all n > k. Let ( xk, xk, xk,…….)  x k . Then x k  W. By (2)
    
d ( x , x k ) = lim d( xn, xk )  2 < . Therefore, B( x , )  W  φ for all  > 0.
n
___ 
Hence W = X .

(c) Proving the completeness of X .
  ___   
Let ( x n ) be any Cauchy sequence in X . Since W = X , then for all x n  X there
    1      
is zn W such that d ( x n , zn ) < n . Then d ( zm , zn )  d ( zm , xm ) +
      1    1 
d ( xm , x n ) + d ( x n , z n ) < m + d ( xm , x n ) + n . However, ( x n ) is a
1   
Cauchy sequence, then for every  > 0, there is k such that m + d ( xm , x n ) +
1 
n < . Therefore, ( zm ) is Cauchy. Since T : X  W is an isometry and
   
zm W, then the sequence ( zm), where zm = T zm is Cauchy in X. Let x  X
-1

          1   
and (zm) x , then d ( x n , x )  d ( x n , zn ) + d ( zn , x ) < n + d ( zn , x ) ..… (3)
  
Since (zm)  x and zn W, then (zn, zn, zn,…….)  zn and (3) becomes
   1 1
d ( xn , x ) < n + lim d( zn, zm ). However, (zm) is Cauchy and ( n )
n

16
converges, then for every  > 0, there is k sufficiently large such that for all n >k,
      
d ( x n , x ) < . Hence ( x n ) converges to x in X . Therefore, X is complete.

(d) Proving the uniqueness of X , except for isometries.
~ ~ ~ ~
Let ( X , d ) be another complete metric space with a subspace W dense in X and
~ ~ ~ ~ ~ ~
isometric to X, then for any x , y  X we have sequences ( xn ), ( yn ) W such that
~ ~ ~ ~ ~ ~ ~ ~ ~ ~
xn  x and yn  y . As in (a) one can show that | d ( x , y ) - d ( xn , yn ) | 
~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~
d ( x , xn ) + d ( y , yn ) 0 as n . Hence d ( x , y ) = lim d ( xn , yn ). Let
n
~  ~ ~
T : W  W  X be a bijection isometry. Since ( xn ), ( yn ) are Cauchy sequences
~ ~ ~ ~ ~ ~
in W that are converges to x and y in X , respectively, then (T xn ) and (T yn ) are
 
Cauchy sequences in W. However, W is dense in X , then they converges in X ,
~  ~  ~
say lim T xn = x and lim T yn = y . Let S be the extension of T to X , and
n n
~  ~  ~  ~ ~
S : X  X defined by S x = x , where lim T xn = x and lim xn = x . By using
n n
~ ~ ~ ~ ~ ~  ~ ~
the isometry of T we have, d ( x , y ) = lim d ( xn , yn ) = lim d ( T xn , T yn ) =
n n
    ~ ~  
d ( x , y ) = d ( S x , S y ). Therefore, S is an isometry and so 1-1. Let x  X , then
  
there is a sequence ( x n ) of elements in W such that lim x n = x . Since T is onto,
n
~ ~ ~ 
then there is a sequence ( xn ) of elements in W such that Then T xn = x n for all n.
~  ~
Then lim T xn = x . So that (T xn ) is a Cauchy sequence in W. However, T is an
n
~ ~ ~ ~
isometry, then ( xn ) = (T -1T xn ) is a Cauchy sequence in W . Since X is
~ ~ ~ ~ ~ 
complete then there is x  X such that xn  x . By the definition of S, S x = x .
~ 
Thus S is onto. Therefore, X and X are isometric. This completes the proof

Note. The space X occurring in the above theorem is called the completion of
the given space X.

H. W. 1, 3-5, 8-13. H. W.* 4, 8.

17

You might also like