MathCity.

org
Merging man and maths

Metric Spaces
Available online @ http://www.mathcity.org/msc, Version: 2.0

v Metric Spaces Let X be a non-empty set and denotes the set of real numbers. A function d : X X is said to be metric if it satisfies the following axioms " x, y, z X . [M1] d ( x, y ) 0 i.e. d is finite and non-negative real valued function. [M2] d ( x, y ) = 0 if and only if x = y. [M3] d ( x, y ) = d ( y, x) (Symmetric property) (Triangular inequality) [M4] d ( x, z ) d ( x, y ) + d ( y, z ) The pair (X, d ) is then called metric space. d is also called distance function and d(x, y) is the distance from x to y. Note: If (X, d) be a metric space then X is called underlying set. v Examples: i) Let X be a non-empty set. Then d : X X defined by if x y 1 d ( x, y ) = if x = y 0 is a metric on X and is called trivial metric or discrete metric. ii) Let be the set of real number. Then d : defined by d ( x, y ) = x - y is a metric on . The space ( , d ) is called real line and d is called usual metric on . iii) Let X be a non-empty set and d : X X be a metric on X. Then d : X X defined by d ( x, y ) = min (1, d ( x, y ) ) is also a metric on X. Proof: [M1] Since d is a metric so d ( x, y ) 0 as d ( x, y ) is either 1 or d ( x, y ) so d ( x, y ) 0 . [M2] If x = y then d ( x, y ) = 0 and then d ( x, y ) which is min (1, d ( x, y ) ) will be zero. Conversely, suppose that d ( x, y ) = 0 min (1, d ( x, y ) ) = 0 d ( x, y ) = 0 x = y as d is metric. [M3] d ( x, y ) = min (1, d ( x, y ) ) = min (1, d ( y, x) ) = d ( y, x) Q d ( x, y ) = d ( y , x ) [M4] We have d ( x, z ) = min (1, d ( x, z ) ) d ( x, z ) 1 or d ( x, z ) d ( x, z ) We wish to prove d ( x, z ) d ( x, y ) + d ( y, z ) now if d ( x, z ) 1 , d ( x, y ) 1 and d ( y, z ) 1 then d ( x, z ) = 1 , d ( x, y ) = 1 and d ( y, z ) = 1 and d ( x, y ) + d ( y, z ) = 1 + 1 = 2

Metric Spaces

therefore d ( x, z ) d ( x, y ) + d ( y, z ) Now if d ( x, z ) < 1 , d ( x, y ) < 1 and d ( y, z ) < 1 Then d ( x, z ) = d ( x, z ) , d ( x, y ) = d ( x, y ) and d ( y, z ) = d ( y, z ) As d is metric therefore d ( x, z ) d ( x, y ) + d ( y, z ) d ( x, z ) d ( x, y ) + d ( y , z ) Q.E.D iv) Let d : X X be a metric space. Then d : X X defined by d ( x, y ) d ( x, y ) = is also a metric. 1 + d ( x, y ) Proof. d ( x, y ) = d ( x, y ) 0 [M1] Since d ( x, y ) 0 therefore 1 + d ( x, y ) d ( x, y ) [M2] Let d ( x, y ) = 0 = 0 d ( x, y ) = 0 x = y 1 + d ( x, y ) Now conversely suppose x = y then d ( x, y ) = 0 . d ( x, y ) 0 Then d ( x, y ) = = =0 1 + d ( x, y ) 1 + 0 d ( x, y ) d ( y, x) = = d ( y, x ) [M3] d ( x, y ) = 1 + d ( x, y ) 1 + d ( y , x ) [M4] Since d is metric therefore d ( x, z ) d ( x, y ) + d ( y, z ) a b Now by using inequality a < b < . 1+ a 1+ b d ( x, z ) d ( x, y ) + d ( y , z ) We get 1 + d ( x, z ) 1 + d ( x, y ) + d ( y , z ) d ( x, y ) d ( y, z ) d ( x, z ) + 1 + d ( x, y ) + d ( y , z ) 1 + d ( x, y ) + d ( y , z ) d ( x, y ) d ( y, z ) d ( x, z ) + 1 + d ( x, y ) 1 + d ( y , z ) d ( x, z ) d ( x, y ) + d ( y , z ) Q.E.D v) The space C[a, b] is a metric space and the metric d is defined by d ( x, y ) = max x(t ) - y (t )
tJ

where J = [a, b] and x, y are continuous real valued function defined on [a, b]. Proof. [M1] Since x(t ) - y (t ) 0 therefore d ( x, y ) 0 . [M2] Let d ( x, y ) = 0 x(t ) - y (t ) = 0 x(t ) = y (t ) Conversely suppose x = y Then d ( x, y ) = max x(t ) - y (t ) = max x(t ) - x(t ) = 0
tJ tJ

Available at http://www.MathCity.org

Metric Spaces

[M3] d ( x, y ) = max x(t ) - y (t ) = max y (t ) - x(t ) = d ( y, x)

tJ tJ

[M4] d ( x, z ) = max x(t ) - z (t ) = max x(t ) - y (t ) + y (t ) - z (t )

tJ tJ tJ

max x(t ) - y (t ) + max y (t ) - z (t )

tJ

= d ( x, y ) + d ( y , z ) Q.E.D vi) d : is a metric, where is the set of real number and d defined by
d ( x, y ) = x- y

vii) Let x = ( x1 , y1 ) , y = ( x2 , y2 ) we define

d ( x, y ) = ( x1 - x2 )2 + ( y1 - y2 ) 2 is a metric on

and called Euclidean metric on 2 or usual metric on 2 . d ( x, y ) = ( x - y ) 2

viii) d : is not a metric, where is the set of real number and d defined by Proof. [M1] Square is always positive therefore ( x - y ) 2 = d ( x, y ) 0 [M2] Let d ( x, y ) = 0 ( x - y ) 2 = 0 x - y = 0 x = y Conversely suppose that x = y then d ( x, y ) = ( x - y )2 = ( x - x)2 = 0 [M3] d ( x, y ) = ( x - y ) 2 = ( y - x) 2 = d ( y, x) [M4] Suppose that triangular inequality holds in d. then for any x, y, z d ( x, z ) d ( x - y ) + d ( y , z ) ( x - z )2 ( x - y ) 2 + ( y - z ) 2 Since x, y, z therefore consider x = 0, y = 1 and z = 2 .

(0 - 2)2 (0 - 1)2 + (1 - 2) 2 4 1+1 42 which is not true so triangular inequality does not hold and d is not metric.

ix) Let x = ( x1 , x2 ) , y = ( y1 , y2 ) 2 . We define d ( x, y ) = x1 - y1 + x2 - y2 is a metric on 2 , called Taxi-Cab metric on 2 . x) Let n be the set of all real n-tuples. For x = ( x1 , x2 ,..., xn ) and y = ( y1 , y2 ,..., yn ) in n we define d ( x, y ) = ( x1 - y1 ) 2 + ( x2 - y2 ) 2 + ... + ( xn - yn )2 then d is metric on n , called Euclidean metric on n or usual metric on n . xi) The space l . As points we take bounded sequence x = ( x1 , x2 ,...) , also written as x = ( xi ) , of complex numbers such that
Available at http://www.MathCity.org

Metric Spaces

xi Cx " i = 1, 2,3,... where C x is fixed real number. The metric is defined as d ( x, y ) = sup xi - yi where y = ( yi )
i

xii) The space l p , p 1 is a real number, we take as member of l p , all sequence

x = (x j ) of complex number such that

x
j =1

p j

< .
1 p

p The metric is defined by d ( x, y ) = x j - h j j =1

Where y = (h j ) such that Proof.

1 p

h
j =1

p j

<

p [M1] Since x j - h j 0 therefore x j - h j = d ( x, y ) 0 . j =1 [M2] If x = y then

p p p p p p d ( x, y ) = x j - h j = x j - x j == 0 = 0 j =1 j =1 j =1 1 1 1

Conversely, if d ( x, y ) = 0 p x j - h j = 0 x j - h j = 0 (x j ) = (h j ) x = y j =1
1 p

p p [M3] d ( x, y ) = x j - h j = h j - x j = d ( y, x ) j =1 j =1

1 p

1 p

[M4] Let z = (z j ) , such that

z
j =1

p j 1 p

<

p then d ( x, z ) = x j - z j j =1

p = x j -h j +hj - z j j =1 * Using Minkowskis Inequality

1 p

1 p

p p x j -h j + h j - z j j =1 j =1 = d ( x, y ) + d ( y , z )

1 p

Available at http://www.MathCity.org

Metric Spaces

5 Q.E.D

v Pseudometric Let X be a non-empty set. A function d : X X is called pseudometric if and only if i) d ( x, x) = 0 for all x X . ii) d ( x, y ) = d ( y, x) for all x, y X . iii) d ( x, z ) d ( x, y ) + d ( y, z ) for all x, y, z X . OR A pseudometric satisfies all axioms of a metric except d ( x, y ) = 0 may not imply x = y but x = y implies d ( x, y ) = 0 . Example Let x, y 2 and x = ( x1 , x2 ) , y = ( y1 , y2 ) Then d ( x, y ) = x1 - y1 is a pseudometric on 2 . Let x = (2,3) and y = (2,5) Then d ( x, y ) = 2 - 2 = 0 but x y

Note: Every metric is a pseudometric, but pseudometric is not metric.

* Minkowskis Inequality If xi = (x1 , x 2 ,..., x n ) and hi = (h1 ,h2 ,...,hn ) are in n and p > 1 , then xi + hi i =1
p

xi i =1

1 p

+ hi i =1

1 p

1 p

v Distance between sets Let ( X , d ) be a metric space and A, B X . The distance between A and B denoted by d ( A, B) is defined as d ( A, B) = inf {d (a, b) | a A, b B} If A = { x} is a singleton subset of X, then d ( A, B) is written as d ( x, B ) and is called distance of point x from the set B. v Theorem Let ( X , d ) be a metric space. Then for any x, y X d ( x, A) - d ( y, A) d ( x, y ) Proof. Let z A then d ( x, z ) d ( x, y ) + d ( y, z ) then d ( x, A) = inf d ( x, z ) d ( x, y ) + inf d ( y, z ) = d ( x, y ) + d ( y, A) d ( x, A) - d ( y, A) d ( x, y ) (i) Next
Available at http://www.MathCity.org
zA zA

6
z A

Metric Spaces

d ( y, A) = inf d ( y, z ) d ( y, x) + inf d ( x, z ) = d ( y, x) + d ( x, A) - d ( x, A) + d ( y, A) d ( y, x) - ( d ( x, A) - d ( y, A) ) d ( x, y ) (ii ) Combining equation (i) and (ii) d ( x, A) - d ( y, A) d ( x, y ) Q.E.D

zA

Q d ( x, y ) = d ( y , x )

v Diameter of a set Let ( X , d ) be a metric space and A X , we define diameter of A denoted by d ( A) = sup d (a, b)
a , b A

Note: For an empty set j , following convention are adopted (i) d (j ) = - , some authors take d (j ) also as 0. (ii) d ( p,j ) = i.e distance of a point p from empty set is . (iii) d ( A, j ) = , where A is any non-empty set.
v Bounded Set Let ( X , d ) be a metric space and A X , we say A is bounded if diameter of A is finite i.e. d ( A) < . v Theorem The union of two bounded set is bounded. Proof. Let ( X , d ) be a metric space and A, B X be bounded. We wish to prove A B is bounded. Let x, y A B If x, y A then since A is bounded therefore d ( x, y ) < and hence d ( A B ) = sup d ( x, y ) < then A B is bounded.
x , y A B

Similarity if x, y B then A B is bounded. Now if x A and y B then d ( x, y ) d ( x, a) + d (a, b) + d (b, y ) Since d ( x, a), d (a, b) and d (b, y ) are finite Therefore d ( x, y ) < i.e A B is bounded.

where a A, b B . Q.E.D

v Open Ball Let ( X , d ) be a metric space. An open ball in ( X , d ) is denoted by B( x0 ; r ) = { x X | d ( x0 , x) < r} x0 is called centre of the ball and r is called radius of ball and r 0 .

Available at http://www.MathCity.org

Metric Spaces

v Closed Ball The set B( x0 ; r ) = { x X | d ( x0 , x) r} is called closed ball in ( X , d ) . v Sphere The set S ( x0 ; r ) = { x X | d ( x0 , x) = r} is called sphere in ( X , d ) . v Examples Consider the set of real numbers with usual metric d = x - y " x, y then B( xo ; r ) = { x | d ( xo , x) < r} i.e. B( xo ; r ) = { x : x - xo < r} i.e. x0 - r < x < x + r = ( x0 - r , x0 + r ) i.e. open ball is the real line with usual metric is an open interval. And B( xo ; r ) = { x : x - x0 r} i.e. x0 - r x x0 + r = [ x0 - r , x0 + r ] i.e. closed ball in a real line is a closed interval. And S ( xo ; r ) = { x : x - x0 = r} = { x0 - r , x0 + r} i.e. two point x0 - r and x0 + r only.

v Open Set Let ( X , d ) be a metric space and set G is called open in X if for every x G , there exists an open ball B( x ; r ) G . v Theorem An open ball in metric space X is open. Proof. Let B( x0 ; r ) be an open ball in ( X , d ) . Let y B ( x0 ; r ) then d ( x0 , y ) = r1 < r Let r2 < r - r1 , then B( y ; r2 ) B ( x0 ; r ) Hence B( x0 ; r ) is an open set.

Alternative:
Let B( x0 ; r ) be an open ball in ( X , d ) . Let x B( x0 ; r ) then d ( x0 , x) = r1 < r Take r2 = r - r1 and consider the open ball B ( x ; r2 ) we show that B( x ; r2 ) B ( x ; r ) . For this let y B ( x ; r2 ) then d ( x, y ) < r2 and d ( x0 , y ) d ( x0 , x) + d ( x, y ) < r1 + r2 = r hence y B ( x0 ; r ) so that B( x; r2 ) B ( x0 ; r ) . Thus B( x0 ; r ) is an open. Q.E.D
Available at http://www.MathCity.org

Metric Spaces

Note: Let ( X , d ) be a metric space then i) X and j are open sets.

ii) Union of any number of open sets is open. iii) Intersection of a finite number of open sets is open. v Limit point of a set Let ( X , d ) be a metric space and A X , then x X is called a limit point or accumulation point of A if for every open ball B( x ; r ) with centre x, B( x ; r ) { A - {x}} j . i.e. every open ball contain a point of A other than x. v Closed Set A subset A of metric space X is closed if it contains every limit point of itself. The set of all limit points of A is called the derived set of A and denoted by A . v Theorem A subset A of a metric space is closed if and only if its complement Ac is open. Proof. Suppose A is closed, we prove Ac is open. Let x Ac then x A . x is not a limit point of A. then by definition of a limit point there exists an open ball B( x ; r ) such that B( x ; r ) A = j . This implies B( x ; r ) Ac . Since x is an arbitrary point of Ac . So Ac is open. Conversely, assume that Ac is an open then we prove A is closed. i.e. A contain all of its limit points. Let x be an accumulation point of A. and suppose x Ac . then there exists an open ball B( x ; r ) Ac B ( x ; r ) A = j . This shows that x is not a limit point of A. this is a contradiction to our assumption. Hence x A . Accordingly A is closed. The proof is complete. v Theorem A closed ball is a closed set. Proof. Let B( x ; r ) be a closed ball. We prove B ( x ; r ) = C (say) is an open ball. Let y C then d ( x, y ) > r . Let r1 = d ( x, y ) then r1 > r . And take r2 = r1 - r r r Consider the open ball B y ; 2 we prove B y ; 2 C . 2 2
Available at http://www.MathCity.org
c

Metric Spaces

r r For this let z B y ; 2 then d ( z, y ) < 2 2 2 By the triangular inequality d ( x, y ) d ( x, z ) + d ( z , y ) d ( x, y ) d ( z , x ) + d ( z , y ) Q d ( y, z ) = d ( z , y ) d ( z , x ) d ( x, y ) - d ( z , y ) 2r - r 2r - r + r r r +r d ( z, x) > r1 - 2 = 1 2 = 1 1 = 1 Q r2 = r1 - r 2 2 2 2 r+r d ( z, x) > =r Q r1 - r = r2 > 0 \ r1 > r 2 z B ( x ; r ) This shows that z C r B y; 2 C 2 Hence C is an open set and consequently B( x ; r ) is closed. Q.E.D v Theorem Let ( X , d ) be a metric space and A X . If x X is a limit point of A. then every open ball B( x ; r ) with centre x contain an infinite numbers of point of A. Proof. Suppose B( x ; r ) contain only a finite number of points of A. Let a1 , a2 ,..., an be those points. and let d ( x, ai ) = ri where i = 1,2,..., n . also consider r = min(r1 , r2 ,..., rn ) Then the open ball B( x ; r ) contain no point of A other than x. then x is not limit point of A. This is a contradiction therefore B( x; r ) must contain infinite numbers of point of A. v Closure of a Set Let ( X , d ) be a metric space and M X . Then closure of M is denoted by M = M M where M is the set of all limit points of M. It is the smallest closed superset of M. v Dense Set Let (X, d) be a metric space the a set M X is called dense in X if M = X . v Countable Set A set A is countable if it is finite or there exists a function f : A which is oneone and onto, where is the set of natural numbers. e.g. , and are countable sets . The set of real numbers, the set of irrational numbers and any interval are not countable sets. v Separable Space A space X is said to be separable if it contains a countable dense subsets.
Available at http://www.MathCity.org

10

Metric Spaces

e.g. the real line is separable since it contain the set of rational numbers, which is dense is . v Theorem Let (X, d) be a metric space, A X is dense if and only if A has non-empty intersection with any open subset of X. Proof. Assume that A is dense in X. then A = X . Suppose there is an open set G X such that A G = j . Then if x G then A ( G - {x}) = j which show that x is not a limit point of A. This implies x A but x X A X This is a contradiction. Consequently A G j for any open G X . Conversely suppose that A G j for any open G X . We prove A = X , for this let x X . If x A then x A A = A then X = A . If x A then let {Gi } be the family of all the open subset of X such that x Gi for every i. Then by hypothesis A Gi j for any i. i.e Gi contain point of A other then x. This implies that x is an accumulation point of A. i.e. x A Accordingly x A A = A and X = A . The proof is complete. v Neighbourhood of a Point Let (X, d) be a metric space and x0 X and a subset N X is called a neighbourhood of x0 if there exists an open ball B( x0 ; e ) with centre x0 such that B( x0 ; e ) N . Shortly neighbourhood is written as nhood . v Interior Point Let (X, d) be a metric space and A X , a point x0 X is called an interior point of A if there is an open ball B( x0 ; r ) with centre x0 such that B( x0 ; r ) A . The set of all interior points of A is called interior of A and is denoted by int(A) or Ao . It is the largest open set contain in A. i.e. Ao A . v Continuity A function f : ( X , d ) (Y , d ) is called continuous at a point x0 X if for any e > 0 there is a d > 0 such that d ( f ( x), f ( x0 ) ) < e for all x satisfying d ( x, x0 ) < d . Alternative: f : X Y is continuous at x0 X if for any e > 0 , there is a d > 0 such that x B( x0 ;d ) f ( x) B ( f ( x0 ); e ) .
Available at http://www.MathCity.org

Metric Spaces

11

v Theorem
f :( X , d ) (Y , d ) is continuous at x0 X if and only if f -1 (G ) is open is X. wherever G is open in Y. Note : Before proving this theorem note that if f : X Y , f -1 : Y X and A X , B Y then f -1 f ( A) A and f f -1 ( B ) B Proof. Assume that f : X Y is continuous and G Y is open. We will prove f -1 (G ) is open in X. Let x f -1 (G ) f ( x) f f -1 (G ) G When G is open, there is an open ball B ( f ( x); e ) G . Since f : X Y is continuous, therefore for e > 0 there is a d > 0 such that

y B ( x ;d ) f ( y ) B ( f ( x); e ) G then y f -1 f (G ) f -1 (G ) Since y is an arbitrary point of B( x ;d ) f -1 (G ) . Also x was arbitrary, this show that f -1 (G ) is open in X. Conversely, for any G Y we prove f : X Y is continuous. For this let x X and e > 0 be given. Now f ( x) Y and let B ( f ( x); e ) be an open ball As y B ( x ;d ) f -1 ( B ( f ( x); e ) )

in Y. then by hypothesis f -1 ( B ( f ( x); e ) ) is open in X and x f -1 ( B ( f ( x); e ) ) f ( y ) f f -1 ( B ( f ( x); e ) ) B ( f ( x); e ) i.e. f ( y ) B ( f ( x); e ) Consequently f : X Y is continuous. The proof is complete.

v Convergence of Sequence: Let ( xn ) = ( x1 , x2 ,...) be a sequence in a metric space ( X , d ) , we say ( xn ) converges to x X if lim d ( xn , x) = 0 .
n

We write lim xn = x or simply xn x as n . Alternatively, we say xn x if for every e > 0 there is an n0 , such that " n > n0 , d ( xn , x) < e . v Theorem If ( xn ) is converges then limit of ( xn ) is unique. Proof. Suppose xn a and xn b , Then 0 d (a, b) d (a, xn ) + d ( xn , b) 0 + 0 as n d ( a, b) = 0 a = b Hence the limit is unique. 8 Alternative Suppose that a sequence ( xn ) converges to two distinct limits a and b. and d (a, b) = r > 0 Since xn a , given any e > 0 , there is a natural number n1 depending on e
Available at http://www.MathCity.org
n

12 such that

Metric Spaces

e whenever n > n1 2 Also xn b , given any e > 0 , there is a natural number n2 depending on e such that e d ( xn , b) < whenever n > n2 2 Take n0 = max(n1 , n2 ) then e e d ( xn , a) < and d ( xn , b) < whenever n > n0 2 2 Since e is arbitrary, take e = r then r = d (a, b) d (a, xn ) + d ( xn , b) r r e < + =r Q d (a, xn ) = d ( xn , a) < 2 2 2 Which is a contradiction, Hence a = b i.e. limit is unique.
d ( xn , a) < v Theorem i) A convergent sequence is bounded. ii) If xn x and yn y then d ( xn , yn ) d ( x, y ) . Proof. (i) Suppose xn x , therefore for any e > 0 there is n0 such that " n > n0 , d ( xn , x) < e Let a = max {d ( x1 , x), d ( x2 , x),............, d ( xn , x)} and k = max {e , a} Then by using triangular inequality for arbitrary xi , x j ( xn )
0 d ( xi , x j ) d ( xi , x) + d ( x, x j )

k + k = 2k Hence ( xn ) is bounded. (ii) By using triangular inequality d ( xn , yn ) d ( xn , x ) + d ( x, y ) + d ( y, yn ) d ( xn , yn ) - d ( x, y ) d ( xn , x ) + d ( y, yn ) 0 + 0 Next d ( x, y ) d ( x, xn ) + d ( xn , yn ) + d ( yn , y ) d ( x, y ) - d ( xn , yn ) d ( x, xn ) + d ( yn , y ) 0 + 0 From (i) and (ii) d ( xn , yn ) - d ( x, y ) 0 as n Hence
lim d ( xn , yn ) = d ( x, y )
n

as n ...........(i ) as n ..........(ii )

Q.E.D

v Cauchy Sequence A sequence ( xn ) in a metric space ( X , d ) is called Cauchy if any e > 0 there is a n0 such that " m, n > n0 , d ( xm , xn ) < e .
Available at http://www.MathCity.org

Metric Spaces

13

v Theorem A convergent sequence in a metric space ( X , d ) is Cauchy. Proof. Let xn x X , therefore any e > 0 there is n0 such that e e " m, n > n0 , d ( xn , x) < and d ( xm , x) < . 2 2 Then by using triangular inequality d ( xm , xn ) d ( xm , x ) + d ( x, xn ) d ( xm , x ) + d ( xn , x ) Q d ( x, y ) = d ( y , x ) e e < + =e 2 2 Thus every convergent sequence in a metric space is Cauchy. v Example Let ( xn ) be a sequence in the discrete space ( X , d ) . If ( xn ) be a Cauchy sequence, then for e = 12 , there is a natural number n0 depending on e such that d ( xm , xn ) < 12 " m, n n0 Since in discrete space d is either 0 or 1 therefore d ( xm , xn ) = 0 xm = xn = x (say) Thus a Cauchy sequence in ( X , d ) become constant after a finite number of terms, i.e.

( xn ) = ( x1, x2 ,..., xn , x, x, x,...)

0

v Subsequence Let (a1 , a2 , a3 ,...) be a sequence ( X , d ) and let (i1 , i2 , i3 ,...) be a sequence of positive integers such that i1 < i2 < i3 < ... then

( an : n ) .

( a , a , a ,...) is called subsequence of

i1 i2 i3

v Theorem (i) Let ( xn ) be a Cauchy sequence in ( X , d ) , then ( xn ) converges to a point x X if and only if ( xn ) has a convergent subsequence xnk which converges to x X . (ii) If ( xn ) converges to x X , then every subsequence xnk also converges to x X . Proof. (i) Suppose xn x X then ( xn ) itself is a subsequence which converges to x X . Conversely, assume that xnk is a subsequence of ( xn ) which converges to x. Then for any e > 0 there is n0 such that " nk > n0 , d xnk , x < Further more ( xn ) is Cauchy sequence

( )

( )

( )

e . 2

e . 2 Suppose n2 = max(n0 , n1 ) then by using the triangular inequality we have

Then for the e > 0 there is n1 such that " m, n > n1 , d ( xm , xn ) <
Available at http://www.MathCity.org

14

Metric Spaces

d ( xn , x ) d xn , xnk + d xnk , x <

) (

)
" nk , n > n2

e e + =e 2 2

This show that xn x . (ii) xn x implies for any e > 0 $ n0 such that d ( xn , x ) < e Then in particular d xnk , x < e Hence xnk x X . v Example Let X = ( 0,1) then ( xn ) = ( x1 , x2 , x3 ,...) = ( 1 2 , 1 3 , 1 4 ,...) is a sequence in X. Then xn 0 but 0 is not a point of X. v Theorem Let ( X , d ) be a metric space and M X . (i) Then x M if and only if there is a sequence ( xn ) in M such that xn x . (ii) If for any sequence ( xn ) in M, xn x x M , then M is closed. Proof. (i) Suppose x M = M M If x M , then there is a sequence ( x, x, x,...) in M which converges to x. If x M , then x M i.e. x is an accumulation point of M, therefore each n the 1 open ball B x ; contain infinite number of point of M. n 1 We choose xn M from each B x ; n 1 Then we obtain a sequence ( xn ) of points of M and since 0 as n . n Then xn x as n . Conversely, suppose ( xn ) such that xn x . We prove x M If x M then x M . Q M = M M If x M , then every neighbourhood of x contain infinite number of terms of ( xn ) . Then x is a limit point of M i.e. x M Hence x M = M M . (ii) If ( xn ) is in M and xn x , then x M then by hypothesis M = M , then M is closed. v Complete Space A metric space ( X , d ) is called complete if every Cauchy sequence in X converges to a point of X.
Available at http://www.MathCity.org

" nk > n0

Metric Spaces

15

v Subspace Let ( X , d ) be a metric space and Y X then Y is called subspace if Y is itself a metric space under the metric d. v Theorem A subspace of a complete metric space ( X , d ) is complete if and only if Y is closed in X. Proof. Assume that Y is complete we prove Y is closed. Let x Y then there is a sequence ( xn ) in Y such that xn x . Since convergent sequence is a Cauchy and Y is complete then xn x Y . Since x was arbitrary point of Y
Y Y

QY Y Therefore Y = Y Consequently Y is closed. Conversely, suppose Y is closed and ( xn ) is a Cauchy sequence. Then ( xn ) is Cauchy in X and since X is complete so xn x X .

Also x Y and Y X . Since Y is closed i.e. Y = Y therefore x Y . Hence Y is complete. 8 v Nested Sequence: A sequence sets A1 , A2 , A3 ,... is called nested if A1 A2 A3 ... v Theorem (Cantors Intersection Theorem) A metric space ( X , d ) is complete if and only if every nested sequence of nonempty closed subset of X, whose diameter tends to zero, has a non-empty intersection. Proof. Suppose ( X , d ) is complete and let A1 A2 A3 ... be a nested sequence of closed subsets of X. Since Ai is non-empty we choose a point an from each An . And then we will prove (a1 , a2 , a3 ,...) is Cauchy in X. Let e > 0 be given, since lim d ( An ) = 0 then there is n0 such that d An0 < 0 Then for m, n > n0 , d (am , an ) < e . This shows that (an ) is Cauchy in X. Since X is complete so an p X (say) Suppose the contrary that p I An then $ a k such that p Ak .
n n

( )

We prove p I An ,
n

Since Ak is closed, d ( p, Ak ) = d > 0 .

Available at http://www.MathCity.org

16

Metric Spaces

d d Consider the open ball B p ; then Ak and B p ; are disjoint 2 2 d Now ak , ak +1 , ak + 2 ,... all belong to Ak then all these points do not belong to B p ; 2 This is a contradiction as p is the limit point of ( an ) . Hence p I An .
n

Conversely, assume that every nested sequence of closed subset of X has a non-empty intersection. Let ( xn ) be Cauchy in X, where ( xn ) = ( x1 , x2 , x3 ,...) Consider the sets A1 = { x1 , x2 , x3 ,...} A2 = { x2 , x3 , x4 ,...} Ak = { xn : n k} Then we have A1 A2 A3 ... We prove lim d ( An ) = 0 Since ( xn ) is Cauchy, therefore $ n0 such that " m, n > n0 , d ( xm , xn ) < e , i.e. Now d An = d ( An ) then lim d ( An ) = lim d ( An ) = 0
n n n

lim d ( An ) = 0 .
n

( )

Also A1 A2 A3 ... Then by hypothesis We prove xn p X

n n

IA

j . Let p I An
n

Since lim d ( An ) = 0 therefore $ k0 such that d Ak0 < e Then for n > k0 , xn , p An d ( xn , p ) < e " n > k0 This proves that xn p X . The proof is complete. v Complete Space (Examples) (i) The discrete space is complete. Since in discrete space a Cauchy sequence becomes constant after finite terms i.e. ( xn ) is Cauchy in discrete space if it is of the form ( x1 , x2 , x3 ,..., xn = b, b, b,...) (ii) The set = {0, 1, 2,...} of integers with usual metric is complete. (iii) The set of rational numbers with usual metric is not complete.
Available at http://www.MathCity.org

( )

Metric Spaces

17 2,

Q (1.1,1.41,1.412,...) is a Cauchy sequence of rational numbers but its limit is which is not rational.

(iv) The space of irrational number with usual metric is not complete. We take ( -1,1) , ( - 1 2 , 1 2 ) , ( - 13 , 13 ) ,..., ( - 1 n , 1 n ) We choose one irrational number from each interval and these irrational tends to zero as we goes toward infinity, as zero is a rational so space of irrational is not complete. v Theorem The real line is complete. Proof. Let ( xn ) be any Cauchy sequence of real numbers. We first prove that ( xn ) is bounded. Let e = 1 > 0 then $ n0 such that " m, n n0 , d ( xm , xn ) = xm - xn < 1 In particular for n n0 we have Let a = max x1 , x2 ,..., xn0 + 1 and b = min x1 , x2 ,..., xn0 - 1

xn0 - xn 1 xn0 - 1 xn xn0 + 1

If the range of the sequence is { xn } = { x1 , x2 , x3 ,...} is finite, then one of the term is the sequence say b will repeat infinitely i.e. b, b ,b ,. Then (b, b, b,...) is a convergent subsequence which converges to b. If the range is infinite then by the Bolzano Weirestrass theorem, the bounded infinite set { xn } has a limit point, say b. Then each of the open interval S1 = (b - 1, b + 1) , S2 = ( b - 1 2 , b + 1 2 ) , S2 = ( b - 13 , b + 13 ) , has an infinite numbers of points of the set { xn } . i.e. there are infinite numbers of terms of the sequence ( xn ) in every open interval Sn . We choose a point xi1 from S1 , then we choose a point xi2 from S2 such that i1 < i2 i.e. the terms xi2 comes after xi1 in the original sequence ( xn ) . Then we choose a term xi3 such that i2 < i3 , continuing in this manner we obtain a subsequence

then b xn a " n . this shows that ( xn ) is bounded with b as lower bound and a as upper bound. Secondly we prove ( xn ) has convergent subsequence ( xni ) .

( x ) = ( x , x , x ,K) .
in i1 i2 i3

It is always possible to choose a term because every interval contain an infinite numbers of terms of the sequence ( xn ) . Since b - 1n b and b + 1n b as n . Hence we have convergent subsequence

( x ) whose limit is
in

b.

Lastly we prove that xn b . Since ( xn ) is a Cauchy therefore for any e > 0 there is n0 such that
Available at http://www.MathCity.org

18
" m, n > n0

Metric Spaces

e 2 Also since xin b there is a natural number im such that im > n0 Then " m, n, im > n0
xm - xn <
d ( xn , b ) = xn - b = xn - xim + xim - b

xn - xim + xim - b < Hence xn b and the proof is complete. v Theorem The Euclidean space n is complete. Proof. Let ( xm ) be any Cauchy sequence in n . Then for any e > 0 , there is n0 such that " m, r > n0

e e + =e 2 2

( m ) ( r ) 2 2 d ( xm , xr ) = x j - x j < e ..............(i) (m) (r ) (m) (m) (m) (m) (r ) (r ) (r ) (r ) where xm = x j = x1 , x 2 , x3 ,K, x n and xr = x j = x1 , x 2 , x3 ,K, x n Squaring both sided of (i) we obtain (m) (r ) x j - x j < e 2
2

xj-xj <e

(m)

(r )

" j = 1, 2,3,K, n

( m ) (1) (2) (3) This implies x j = x j , x j , x j ,K is a Cauchy sequence of real numbers for every j = 1, 2,3,K, n . Since is complete therefore x j x j (say) Using these n limits we define x = (x j ) = (x1 , x 2 , x 3 ,K, x n ) then clearly x n . In (i) as r , d ( xm , x ) < e " m > n0 which show that xm x n And the proof is complete. Note: In the above theorem if we take n = 2 then we see complex plane = 2 is complete. Moreover the unitary space n is complete. v Theorem The space l is complete. Proof.
Available at http://www.MathCity.org
(m)

We prove xm x

Metric Spaces

19

Let ( xm ) be any Cauchy sequence in l . Then for any e > 0 there is n0 > such that " m, n > n0 d ( xm , xn ) = sup x j - x j < e .......... (i)
j (m) n

(m) (m) (m) (m) (n) (n) (n) (n) Where xm = x j = x1 , x 2 , x3 ,K and xn = x j = x1 , x 2 , x3 ,K Then from (i)

x j - x j < e ..........(ii)

(m)

(n)

" j = 1, 2,3,K and " m, n > n0

( m ) (1) (2) (3) It means x j = x j , x j , x j ,K is a Cauchy sequence of real or complex numbers for every j = 1, 2,3,K Using these infinitely many limits we define x = (x j ) = (x1 , x 2 , x3 ,K) . We prove x l and xm x . In (i) as n we obtain x j - x j < e ..........(iii ) " m > n0 We prove x is bounded. By using the triangular inequality
(m)

And since and are complete therefore x j x j or (say).

(m)

x j = x j - x j + x j x j - x j + x j < e + km
Where x j < km as xm is bounded. Hence (x j ) = x is bounded. This shows that xn x l . And the proof is complete. v Theorem The space C of all convergent sequence of complex number is complete. Note: It is subspace of l . Proof. First we prove C is closed in l . Let x = (x j ) C , then there is a sequence ( xn ) in C such that xn x , (n) (n) (n) (n) where xn = x j = x1 , x 2 , x3 ,K . Then for any e > 0 , there is n0 such that " n n0 d ( xn , x ) = sup x j - x j <
j (n) (m)

(m)

(m)

(m)

(m)

e 3
Available at http://www.MathCity.org

20

Metric Spaces

such that " j , k > n1

Then in particular for n = n0 and " j = 1,2,3,........... ( n0 ) e x j -xj < 3 Now xn0 C then xn0 is a convergent sequence therefore $ n1

e 3 Then by using triangular inequality we have

( n0 )

x j - xk <
( n0 ) ( n0 )

( n0 )

x j - xk = x j - x j + x j - xk + xk - xk
x j - x j + x j - xk + xk - xk
( n0 ) ( n0 ) ( n0 ) ( n0 )

( n0 )

( n0 )

e e e + + =e " j , k > n1 3 3 3 Hence x is Cauchy in l and x is convergent Therefore x C and C = C . i.e. C is closed in l and l is complete. Since we know that a subspace of complete space is complete if and only if it is closed in the space. Consequently C is complete.
< v Theorem The space l p , p 1 is a real number, is complete. Proof. Let ( xn ) be any Cauchy sequence in l p . Then for every e > 0 , there is n0 such that " m, n > n0 (m) (n) d ( xm , xn ) = x j - x j j =1 (m) (m) (m) (m) where xm = x j = x1 , x 2 , x3 ,K Then from (i)
p

p < e .. (i)

x j - x j < e (ii)

(m)

(n)

" m, n > n0 and for any fixed j.

(m) This shows that x j is a Cauchy sequence of numbers for the fixed j. Using these infinite many limits we define x = (x j ) = (x1 , x 2 , x3 ,K) . We prove x l p and xm x as m . From (i) we have
Available at http://www.MathCity.org

Since and are complete therefore x j x j or (say) as m .

(m)

Metric Spaces

21

k (m) (n) x j -x j j =1 i.e.

p <e < e p . (iii)

x
j =1

(m) j

-x j

(n) p

Taking as n , we get

x
j =1

(m) j

-x j
p

<ep ,

k = 1, 2, 3, .

Now taking k , we obtain

(m) j

-x j

< e p (iv)

" j = 1,2,3,..........

This shows that ( xm - x ) l p Now l p is a vector space and xm l p , x - xm l p then xm + ( x - xm ) = x l p . Also from (iv) we see that p " m > n0 ( d ( xm , x) ) < e p i.e. d ( xm , x) < e " m > n0 This shows that xm x l p as x . And the proof is complete. v Theorem The space C[a, b] is complete. Proof. Let ( xn ) be a Cauchy sequence in C[a, b]. Therefore for every e > 0 , there is n0 such that " m, n > n0 d ( xm , xn ) = max xm (t ) - xn (t ) < e (i) Then for any fix t = t0 J
tJ

where J = [ a, b] .

xm (t0 ) - xn (t0 ) < e " m, n > n0 It means ( x1 (t0 ), x2 (t0 ), x3 (t0 ),K) is a Cauchy sequence of real numbers. And since is complete therefore xm (t0 ) x(t0 ) (say) as m . In this way for every t J , we can associate a unique real number x(t ) with xn (t ) . This defines a function x(t ) on J. We prove x(t ) C[a, b] and xm (t ) x(t ) as m . From (i) we see that xm (t ) - xn (t ) < e for every t J and " m, n > n0 . Letting n , we obtain for all t J xm (t ) - x(t ) < e " m < n0 . Since the convergence is uniform and the xn s are continuous, the limit function x(t ) is continuous, as it is well known from the calculus. Then x(t ) is continuous.
Available at http://www.MathCity.org

22

Metric Spaces

Hence x(t ) C[a, b] , also xm (t ) - x(t ) < e Therefore xm (t ) x(t ) C[a, b] . The proof is complete.

as m

v Theorem If ( X , d1 ) and (Y , d 2 ) are complete then X Y is complete. where x = (x1 ,h1 ) , y = (x 2 ,h2 ) and x1 ,x 2 X , h1 ,h2 Y . Proof. Let ( xn ) be a Cauchy sequence in X Y . Then for any e > 0 , there is n0 such that " m, n > n0

Note: The metric d (say) on X Y is defined as d ( x, y ) = max ( d1 (x1 ,x 2 ) , d 2 (h1 ,h2 ) )

(m) (n) (m) (n) d ( xm , xn ) = max d1 x , x , d 2 h , h < e (m) (n) (m) (n) d1 x , x < e and d 2 h , h < e " m, n > n0 (m) (1) (2) (3) This implies x = x , x , x ,K is a Cauchy sequence in X. ( m ) (1) (2) (3) and h = h , h , h ,K is a Cauchy sequence in Y. Since X and Y are complete therefore x x X (say) and h h Y (say) Let x = (x , m ) then x X Y . Also (m) (m) d ( xm , x ) = max d1 x ,x , d 2 h ,h 0 as n . Hence xm x X Y . This proves completeness of X Y .
(m) (m)

v Theorem f : ( X , d ) (Y , d ) is continuous at x0 X if and only if xn x implies f ( xn ) f ( x0 ) . Proof. Assume that f is continuous at x0 X then for given e > 0 there is a d > 0 such that d ( x, x0 ) < d d ( f ( x), f ( x0 ) ) < e . Let xn x0 , then for our d > 0 there is n0 such that d ( xn , x0 ) < d , " n > n0 Then by hypothesis d ( f ( xn ), f ( x0 ) ) < e , " n > n0 i.e. f ( xn ) f ( x0 ) Conversely, assume that xn x0 f ( xn ) f ( x0 ) We prove f : X Y is continuous at x0 X , suppose this is false
Available at http://www.MathCity.org

Metric Spaces

23

Then there is an e > 0 such that for every d > 0 there is an x X such that d ( x, x0 ) < d but d ( f ( x), f ( x0 ) ) e 1 In particular when d = , there is xn X such that n d ( xn , x0 ) < d but d ( f ( xn ), f ( x0 ) ) e . This shows that xn x0 but f ( xn ) f ( x0 ) as n . This is a contradiction. Consequently f : X Y is continuous at x0 X . The proof is complete. v Rare (or nowhere dense in X ) Let X be a metric, a subset M X is called rare (or nowhere dense in X ) if M has no interior point i.e. int M = j .

( )

v Meager ( or of the first category) Let X be a metric, a subset M X is called meager (or of the first category) if M can be expressed as a union of countably many rare subset of X. v Non-meager ( or of the second category) Let X be a metric, a subset M X is called non-meager (or of the second category) if it is not meager (of the first category) in X. v Example: Consider the set of rationales as a subset of a real line . Let q , then {q} = {q} because - {q} = ( -, q ) ( q, ) is open. Clearly {q} contain no open ball. Hence is nowhere dense in as well as in . Also since is countable, it is the countable union of subsets {q} , q . Thus is of the first category. v Bairs Category Theorem If X j is complete then it is non-meager in itself. OR A complete metric space is of second category. Proof. Suppose that X is meager in itself then X = U M k , where each M k is rare in X.
k =1

Since M 1 is rare then int( M ) = M = j i.e. M 1 has non-empty open subset But X has a non-empty open subset ( i.e. X itself ) then M 1 X . This implies M 1 = X - M 1 is a non-empty and open. We choose a point p1 M 1
c c

c and an open ball B1 = B ( p1; e1 ) M 1 , where e1 < 1 . 2

Available at http://www.MathCity.org

24 Now M 2 is non-empty and open

c

Metric Spaces

c c 1 Then $ a point p2 M 2 and open ball B2 = B ( p2 ; e 2 ) M 2 B p1; e1 2 c 1 ( M 2 has no non-empty open subset then M 2 B p1; e1 is non-empty and open.) 2 c 1 So we have chosen a point p2 from the set M 2 B p1; e1 and an open ball 2 1 1 1 B ( p2 , e 2 ) around it, where e 2 < e1 < < 2-1 . 2 2 2 Proceeding in this way we obtain a sequence of balls Bk such that 1 Bk +1 B pk ; e k Bk where Bk = B ( pk ; e k ) " k = 1, 2,3,....... 2 Then the sequence of centres pk is such that for m > n 1 1 d ( pm , pn ) < e m < m +1 0 as m . 2 2 Hence the sequence ( pk ) is Cauchy. Since X is complete therefore pk p X (say) as k . Also d ( pm , p ) d ( pm , pn ) + d ( pn , p ) 1 < e m + d ( pn , p ) 2 < e m + d ( pn , p ) e m + 0 as n .

p Bm

" m
c

i.e. p M m

" m

Q Bm = M 2 B pm-1; 1 e m-1 2
c

Bm M m Bm M m = j p Mm " m p X This is a contradiction. Bairs Theorem is proof.

References: (1) Lectures (2003-04) Prof. Muhammad Ashfaq Ex Chairman, Department of Mathematics. University of Sargodha, Sargodha. (2) Book Introductory Functional Analysis with Applications By Erwin Kreyszig (John Wiley & Sons. Inc., 1989.)

These notes are available online at http://www.mathcity.org in PDF Format. Last update: July 01, 2011.

Available at http://www.MathCity.org