SEC.

7 COMPLETE METRIC SPACES 1

7. Complete Metric Spaces

7.1. Definitions and examples. The reader is presumably already familiar

with the notion of the completeness of the real line. The real line is, of course,
a particularly simple example of a metric space. We now make the natural
generalization of the notion of completeness to the case of an arbitrary metric
space.

Definition 1. A sequence {xn} of points in a metric space R with

metric ρ is said to satisfy the Cauchy criterion if, given any ε > 0,
there is an integer N ε such that ρ ( x n , x n' ) for all n, n' > N ε .

Definition 2. A subsequence{xn} of points in a metric space R is

called a Cauchy sequence (or a fundamental sequence) if it satisfies
the Cauchy criterion.
Theorem 1. Every convergent sequence {xn} is fundamental.
Proof If{xn} converges to a limit x, then, given any ε > 0, there is
an integer N ε such that

ε
for all n> N ε . But then ρ ( xn , x )<
2

ρ ( x h , x n ) ≤ ρ ( x n , x ) + ρ ( x n' , x ) < ε
'

for all n, n' > N ε .

Definition 3. A metric space R is said to be complete if every Cauchy

sequence in R converges to an element of R. Otherwise R is said to be
incomplete.

Example 1. Let R be the "space of isolated points" considered in Example

1, p. 38. Then the Cauchy sequences in R are just the "stationary sequences,"
i.e., the sequences {xn} all of whose terms are the same starting from some
index n. Every such sequence is obviously convergent to an element of R.
Hence R is complete.
Example 2. The completeness of the real line R1 is familiar from
elementary analysis.

Example 3. The completeness of Euclidean n-space Rn follows from that

of R1. In fact, let
 2 METRIC SPACES CHAP. 2

x( p) =( x (P)
1 , … , x n ) ( p=1 , 2 , … )
( p)

n
be a fundamental sequence of points of Rn. Then, given any ε > 0, there
exists an N ε such that
n

∑ ( x(kp) −x(q)k ) < ε2

2

k =1

for all p, q > N ε . I t follows that

|x (pk )−x(qk )|<ε (k = 1, …, n)

( p)
for all p, q > N ε , i.e., each { x k
} is a fundamental sequence in R1.
Let
x = (x1, …, xn),

where

p
x k = lim x k
P →∞

Then obviously

p
lim x k =x .
P →∞

n
This proves the completeness of Rn. The completeness of the spaces R0 and
n
R1 introduced in Examples 4 and 5, p. 39 is proved in almost the same way
(give the details).
Example 4. Let {xn(t)} be a Cauchy sequence in the function space
c [ a , b ] considered in Example 6, p. 39. Then, given any ε > 0, there is an
N ε such that
|x n ( t ) −x n ( t )|< ε' (1)

for all n , n' > N ε and all t ∈ [ a , b ] . It follows that the sequence { x n (t ) }is
uniformly convergent. But the limit of a uniformly convergent sequence of
continuous functions is itself a continuous function (see Problem 1). Taking
the limit as n' → ∞ in (1), we find that

|x n ( t ) −x ( t )| ≤ ε
for all n > N ε and all t ∈ [ a , b ] , i.e., { x n (t ) } converges in the metric of
c [ a , b ] to a function x (t ) ∈C [ a , b] . Hence C [ a , b]is a complete metric space.
 SEC. 7 COMPLETE METRIC SPACES 3
Example 5. Next let x(n) be a sequence in the space l2 considered in
Example 7, p. 39, so that

x(n) =( x (n1 ) , x(n)

2 , … , xk , …)
(n)

∑ ( x(n)k ) < ∞
2
( n=1 ,2 , … )
k =1
 4 METRIC SPACES CHAP. 2

Suppose further that {x(n)} is a Cauchy sequence. Then, given any ε > 0,
there is a N ε such that

∞
( x n , x n )=∑ (x ¿ ¿ k (n)−x (kn ))<¿ ε ¿ ¿
' '
2
ρ ( 2)
k =1

if n , n' > N ε . It follows that

( x(n)k −x (kn ) ) <ε ( k =1 , 2, … )
' 2

(n)
i.e., for every k the sequence { x k } is fundamental and hence
convergent. Let
(n)
x k =lim x k
n→∞

x=( x 1 , x 2 , … , x k , … )
Then, as we now show, x is itself a point of l2 and moreover {x(n)}
converges to x in the l2 metric, so that l2 is a complete metric space.
In fact, (2) implies

M
2
∑ ( x(n)k −x(kn ) ) <¿ ε ¿
'

(3)
k =1

for any fixed M. Holding n fixed in (3) and taking the limit as n' → ∞,
we get
M
2
∑ ( x(n)k −x(kn ) ) ≤ ε
'

(4)
k =1

Since (4) holds for arbitrary M, we can in turn take the limit of (4) as
M → ∞ , obtaining
∞
2
∑ ( x(n)k −x(kn ) ) ≤ ε
'

(5)
k =1
Just as on p. 40, the convergence of the two series
∞ ∞

∑ ( x ) , ∑ ( x (n)
(n) 2 2
k k −x k )
k =1 k=1

implies the series

 SEC. 7 COMPLETE METRIC SPACES 5
∞

∑ x 2k
k =1

This prove that x ∈ l 2. Moreover, since ε is arbitrary small, (5) implies

√∑ (
∞
2
lim ρ ( x , x )=lim(n)
x (nk )−x k ) =¿ 0
n→∞ n→∞ k=1

i.e., {x(n)} converges to x in the l2 metric, as asserted.

2
Example 6. It is easy to show that the space C [ a, b] of Example 8, p. 40 is
incomplete. If

{
1
−1 if −1≤ t ≤−
n
1 1
φ n ( t )= nt if − ≤t ≤
n n
1
1if ≤t ≤1
n
then { φn ( t ) } is a fundamental sequence in C [−1 1 ] since
2
,
1

∫ ¿¿
−1

However, { φn ( t ) } cannot converge to a function in C [−1 1 ]. In fact,

2
,

consider the discontinuous function

ψ (t)= {−11ifift t≥<00

2
Then, given any function f∈C [ −1 1] , it follows from Schwar’s inequality
,

(obviously still valid for piecewise continuous functions) that

(∫ [ ) (∫ [ ) (∫ [ )
1 1 1 1 1 1
2 2 2 2
f ( t )−ψ (t) ] f ( t )−φn ( t ) ] dt + φn (t )−ψ (t) ] dt
2 2
≤ But
−1 −1 −1
the integral on the left is nonzero, by the continuity of f, and moreover it is clear that

1 zero as n → ∞
lim ∫ [ φ n ( t )−ψ ( t ) ] ⅆt =0cannot converge to
2

n → ∞ −1
 6 METRIC SPACES CHAP. 2

7.2. The nested sphere theorem. A sequence of closed spheres

S[x1, r1], S[x2, r2], …, S[xn, rn], …
in a metric space R is said to be nested (or decreasing) if
S[x1, r1]⊃ S[x2, r2] ⊃…⊃ S[xn, rn] ⊃ …
Using this concept, we can prove a simple criterion for the completeness
of R:
Theorem 2 (Nested sphere theorem). A metric space R is complete if and
only if every nested sequence {Sn} = {S[xn, rn]} of closed spheres in R
such that r n → 0 as n→ 0 has a nonempty intersection

¿ n=1¿ ∞ sn
Proof. If R is complete and if {S n} = {S[xn, rn]} is any nested sequence of
closed spheres in R such that r n → 0 as n→ 0, then the sequence {xn} of
centers of the spheres is fundamental, since ρ ( x n , x n' ) <r n for n' > n and
r n → 0 as n→ 0. Therefore {xn} has a limit. Let

x=lim x n Then xϵ ¿ n=1 ¿ ∞ s n

n→∞

In fact, Sn contains every point of the sequence {xn} except possibly the
points x1, x2, …, xn-1, and hence x is a limit point of every sphere
Sn. But Sn is closed, and hence x ∈ s n for all n.
Conversely, suppose every nested sequence of closed spheres in R
with radii converging to zero has a nonempty intersection, and let {xn}
be any fundamental sequence in R. Then x has a limit in R. To see this,
use the fact that {xn} is fundamental to choose a term x n1 of the
sequence {xn} such that
1
ρ ( xn , xn )<
1
2
for all n ≥ n1, and let S1 be the closed sphere of radius 1 with center x n1 .
Then choose a term x n of {xn} such that n2 > n1 and
2

1
ρ ( xn , xn )< 2
2
2

for all n ≥ n2, and let S2 be the closed sphere of radius ½ with center x n2.
Continue this construction indefinitely, i.e., once having chosen terms
x n1 , x n2 , … , x nk (n1 < n2 < …<nk ), choose a term x nk+ 1 such that x nk+ 1> nk
and
 SEC. 7 COMPLETE METRIC SPACES 7

1
ρ ( xn , xn )< k +1
2
k+ 1

for all n≥nk+1, let Sk+1 be the closed sphere of radius 1 /2k with center
x nk+ 1, and so on. This gives a nested sequence {Sn} of closed spheres
with radii converging to zero. By hypothesis, these spheres have a
nonempty intersection, i.e., there is a point x in all the spheres. This
point is obviously the limit of the sequence { x nk} But if a fundamental
sequence contains a subsequence converging to x, then the sequence
itself must converge to x (why?), i.e.,
lim x n= x
n →∞
7.3. Baire's theorem. It will be recalled from Sec. 6.3 that a subset A of a
metric space R is said to be nowhere dense in R if it is dense in no (open)
sphere at all, or equivalently, if every sphere S ⊂ R contains another sphere
' '
S such that S ∩ A= ∅(check the equivalence). This concept plays an
important role in
Theorem 3 (Baire). A complete metric space R cannot be
represented as the union of a countable number of nowhere dense
sets.
Proof Suppose to the contrary that
R=¿ n=1 ¿ ∞ A n
where every set An is nowhere dense in R. Let S0 ⊂ R be a closed sphere of
radius 1. Since A1 is nowhere dense in S0, being nowhere dense in R, there is a
1
closed sphere S1 of radius less than such that S1 ⊂ S 0 and S1 ∩ A 1= ∅.
3
Since A2 is nowhere dense in S1, being nowhere dense in S0, there is a closed
1
sphere S2 of radius less than such that S2 ⊂ S 1 and S2 ∩ A 2= ∅, and so on.
3
In this way, we get a nested sequence of closed spheres {Sn} with radii
converging to zero such that
Sn ∩ An= ∅ (n=1 , 2 , …)
By the nested sphere theorem, the intersection
¿ n=1¿ ∞ Sn
contains a point x. By construction, x cannot belong to any of the sets
An, i.e.,

x ∉ ¿ n=1¿ ∞ A n It follows that R ∉¿ n=1 ¿ ∞ A n

 8 METRIC SPACES CHAP. 2

contrary to (6). Hence the representation (6) is impossible.

Corollary. A complete metric space R without isolated points is
uncountable.
Proof. Every single-element set {x} is nowhere dense in R.
 SEC. 7 COMPLETE METRIC SPACES 9

7.4. Completion of a metric space. As we now show, an incomplete metric

space can always be enlarged (in an essentially unique way) to give a
complete metric space.
Definition 4. Given a metric space R with closure [R], a
complete metric space R* is called a completion of R if R⊂R*
and [R] = R*, i.e., if R is a subset of R* everywhere dense in
R*.

Example 1. Clearly R* = R if R is already complete (see Problem 7).

Example 2. The space of all real numbers is the completion of the
space of all rational numbers.

Theorem 4. Every metric space R has a completion. This

completion is unique to within an isometric mapping carrying every
point x E R into itself
Proof. The proof is somewhat lengthy, but completely straight
forward. First we prove the uniqueness, showing that if R* and
R** are two completions of R, then there is a one-to-one mapping
¿
x**=φ ( x ) of R* onto R** such that φ (x)=x for all x ∈ R and

ρ1 ( x ¿ , y ¿ )=ρ2 ¿(7)
(y** =φ (y*)), where ρ1 is the distance in R* and ρ2 the distance in
R**. The required mapping φ is constructed as follows: Let x* be an
arbitrary point of R*. Then, by the definition of a completion, there is a
sequence
{xn} of points of R converging to x*. The points of the sequence {xn}
also belong to R**, where they form a fundamental sequence (why?).
Therefore {xn} converges to a point x**∈R**, since R** is complete.
It is clear that x** is independent of the choice of the sequence {xn}
¿
converging to the point x* (why?). If we set φ ( x )= x**, then φ is
the required mapping. In fact, φ ( x )=x for all x∈R, since if
x n → x ∈ R, then obviously x = x* ∈R*, x* * = x. Moreover,
¿ ¿ ¿∗¿ ¿ ¿∗¿¿
suppose x n → x , y n → y in R*, while x n → x , yn → y in
R**. Then, if ρ is the distance in R,
ρ1 ( x ¿ , y ¿ )=lim ρ1 ( x n , y n ) =lim ρ ( xn , y n ) (8)
n→∞ n →∞

(see Problem 3, p. 54), while at the same time

ρ2 ¿ (8' )
But (8) and (8') together imply (7).
 10 METRIC SPACES CHAP. 2

We must now prove the existence of a completion of R. Given an arbitrary

metric space R, we say that two Cauchy sequences {xn } and
 SEC. 7 COMPLETE METRIC SPACES 11

{~
x n} in R are equivalent and write { x n } {~x n } if
lim ρ ( x n , ~
x n )=0
n→∞

As anticipated by the notation and terminology, is reflexive,

symmetric and transitive, i.e., is an equivalence relation in
the sense of Sec. 1.4.Therefore the set of all Cauchy sequences of points in
the space R can be partitioned into classes of equivalent sequences. Let
these classes be the points of a new space R*. Then we define the distance
between two arbitrary points x*, y*∈R* by the formula

ρ1 ( x ¿ , y ¿ )=lim ρ ( x n , y n ) (9)
n→∞
where {xn} is any "representative" of x* (namely, any Cauchy sequence
in the class x*) and {y n} is any representative of y*.
The next step is to verify that (9) is in fact a distance, i.e., that (9)
¿ ¿
exists, is independent of the choice of the sequences { x n } ∈ x , { y n } ∈ y ,
and satisfies the three properties of a distance figuring in Definition 1,
p; 37. Given any ε > 0, it follows from the triangle inequality in R
(recall Problem 1b, p. 45) that
|ρ ( x n , y n )−ρ ( x n , y n )|=¿
' '

|ρ ( x n , y n )−ρ ( x n , y n ) + ρ ( xn , y n) −ρ ( x n , y n )|≤|ρ ( x n , y n ) −ρ ( x n , y n )|+| ρ ( xn , y

' ' ' ' ' '

for all sufficiently large n and n'. Therefore the sequence of real numbers
{Sn} ={ ρ (xn,yn)} is fundamental and hence has a limit. This limit is
¿ ¿
independent of the choice { x n } ∈ x , { y n } ∈ y . In fact, suppose
{ x n } , {~x n }∈ x , { y n } , {~y n } ∈ y .
¿ ¿

Then
|ρ ( x n , y n )−ρ ( ~x n , ~y n)|≤ ρ ( x n , ~x n ) + ρ ( y n , ~y n ) ,
by a calculation analogous to (10). But
lim ρ ( x n , ~
x n )= lim ρ ( y n , ~y n )=0
n→∞ n→ ∞
since { x n } {~x n }, { y n } { ~y n } ,and hence
lim ρ ( x n , y n ) =lim ρ ( ~xn , ~y n )
n→∞ n →∞
As for the three properties of a metric, it is obvious that
ρ1 ( x ¿ , y ¿ )=ρ1 ( y ¿ , x ¿ ), and the fact that ρ1 ( x ¿ , y ¿ )=0 if and only if x*
= y* is an
 12 METRIC SPACES CHAP. 2

immediate consequence of the definition of equivalent Cauchy sequences.

To verify the triangle inequality in R*, we start from the triangle
inequality ρ ( x n , z n ) ≤ ρ ( x n , y n ) + ρ ( y n , z n )
in the original space R and then take the limit as n → ∞ , obtaining
lim ρ ( x n , z n ) ≤ lim ρ¿ ( x¿n , y n )+ lim ρ ( yn , zn )
n→∞ nρ→ (∞x , z ) ≤ ρ ( xn¿→∞
, y ¿ )+ ρ1 ( y ¿ , z ¿ ).
1 1
i.e.,
We now come to the crucial step of showing that R* is a completion of R.
Suppose that with every point x ER, we associate the class x*∈R* of all
Cauchy sequences converging to x. Let
x=lim x n y=lim y n
n →∞ n→∞

Then clearly
ρ ( x , y ) =lim ρ ( x n , y n )
n→ ∞
(recall Problem 3, p. 54), while on the other hand
ρ1 ( x ¿ , y ¿ )=lim ρ ( x n , y n )
n→∞
by definition. Therefore
ρ ( x , y ) =ρ1 ( x ¿ , y ¿ )
and hence the mapping of R into R* carrying x into x* is isometric.
Accordingly, we need no longer distinguish between the original space R
and its image in R*, in particular between the two metrics ρ and ρ1
(recall the relevant comments on p. 44). In other words, R can be regarded
as a subset of R*. The theorem will be proved once we succeed in
showing that
1) R is everywhere dense in R*, i.e., [R] = R;
2) R* is complete.
To this end, given any point x*∈R* and any ε > 0, choose a representative
of x*, namely a Cauchy sequence {xn} in the class x*. Let N be such
that ρ ( x n , x n' ) < ε for all n, n' > N . Then
ρ ( x n , x ¿ )= lim ρ ( x n , x n ) ≤ ε
'
'
n →∞

if n > N, i.e., every neighborhood of the point x* contains a point of R.

It follows that [R] = R.
Finally, to show that R* is complete, we first note that by the very
definition of R*, any Cauchy sequence {xn} consisting of points in R
converges to some point in R*, namely to the point x*∈R* defined by
{xn}. Moreover, since R is dense in R*, given any Cauchy sequence
 SEC. 7 COMPLETE METRIC SPACES 13

¿
{ x n} consisting of points in R*, we can find an equivalent sequence {xn}
consisting of points in R. In fact, we need only choose xn to be any
¿
point of R such that ρ ( x n , x ) < 1/n. The resulting sequence {xn} is
fundamental, and, as just shown, converges to a point x*∈R*. But then the
sequence {xn} also converges to x*.

Example. If R is the space of all rational numbers, then R* is the space

¿
of all real numbers, both equipped with the distance ρ ( x n , x )=|x− y|. In
this way, we can "construct the real number system." However, there still
remains the problem of suitably defining sums and products of real numbers
and verifying that the usual axioms of arithmetic are satisfied.
Problem 1. Prove that the limit f(t) of a uniformly convergent sequence
of functions {fn(t)} continuous on [ a , b ] is itself a function continuous
on [a,b].

Hint. Clearly

|f ( t )−f (t 0 )|≤|f ( t ) −f n ( t )|+|f n ( t )−f n ( t 0)|+|f n ( t0 ) −f ( t 0 )|,

where t,t0∈[a , b]. Use the uniform convergence to make the sum of the
first and third terms on the right small for sufficiently large n. Then use the
continuity of fn(t) to make the second term small for t sufficiently close to
t0
Problem 2. Prove that the space m in Example 9, p. 41 is complete.
Problem 3. Prove that if R is complete, then the intersection
¿ n=1¿ ∞ snfiguring in Theorem 2 consists of a single point.
Problem 4. By the diameter of a subset A of a metric space R is meant the
number
ⅆ ( A ) =¿ x , yϵA ρ ( x , y )
Suppose R is complete, and let {An} be a sequence of closed subsets of
nested in the that
A1 ⊃ A 2 ⊃ … ⊃ A n ⊃ …
Problem 5. A subset A of a metric space R is said to be bounded if its
diameter d(A) is finite. Prove that the union of a finite number of bounded
sets is bounded.
 14 METRIC SPACES CHAP. 2

Problem 6. Give an example of a complete metric space R and a nested

sequence {An} of closed subsets of R such that
¿ n=1¿ ∞ An =∅

Reconcile this example with Problem 4.

Problem 7. Prove that a subspace of a complete metric space R is
complete if and only if it is closed.
Problem 8. Prove that the real line equipped with the distance
ρ ( x , y ) =|arctan x−arctan y|is an incomplete metric space.
Problem 9. Give an example of a complete metric space homeomorphic
to an incomplete metric space.
Hint. Consider the example on p. 44.
Comment. Thus homeomorphic metric spaces can have different "metric
properties."
Problem 10. Carry out the program discussed in the last sentence of the
example on p. 65.
Hint. If {xn} and {yn} are Cauchy sequences of rational numbers serving
as "representatives" of real numbers x* and y*, respectively, define x* + y*
as the real number with representative {xn +yn}.