43.

Complete Metric Spaces 1

Chapter 7. Complete Metric Spaces and

Function Spaces

Note. Recall from your Analysis 1 (MATH 4217/5217) class that the real numbers
R are a “complete ordered field” (in fact, up to isomorphism there is only one such
structure; see my online notes at http://faculty.etsu.edu/gardnerr/4217/
notes/1-3.pdf). The Axiom of Completeness in this setting requires that ev-
ery set of real numbers with an upper bound have a least upper bound. But this
idea (which dates from the mid 19th century and the work of Richard Dedekind)
depends on the ordering of R (as evidenced by the use of the terms “upper” and
“least”). In a metric space, there is no such ordering and so the completeness
idea (which is fundamental to all of analysis) must be dealt with in an alternate
way. Munkres makes a nice comment on page 263 declaring that“completeness is
a metric property rather than a topological one.”

Section 43. Complete Metric Spaces

Note. In this section, we define Cauchy sequences and use them to define complete-
ness. The motivation for these ideas comes from the fact that a sequence of real
numbers is Cauchy if and only if it is convergent (see my online notes for Analysis 1
[MATH 4217/5217] http://faculty.etsu.edu/gardnerr/4217/notes/2-3.pdf;
notice Exercise 2.3.13).
43. Complete Metric Spaces 2

Definition. Let (X, d) be a metric space. A sequence (xn ) of points of X is a

Cauchy sequence on (X, d) if for all ε > 0 there is N ∈ N such that if m, n ≥ N
then d(xn , xm ) < ε. The metric space (X, d) is complete if every Cauchy sequence
in X converges.

Note. By the Triangle Inequality for any metric, a convergent sequence is always
Cauchy (whether the space is complete or not). If the space (X, d) is complete,
then for A a closed subset of X, the subspace (A, d|A ) is complete.

Note. It is not immediately clear how to address Cauchy sequences in a topological

space. Convergence is easy since we have a limit to which to “anchor” open sets, but
this is not the case when addressing Cauchy-ness. John von Neumann published
“On Complete Topological Spaces” in 1935 (Transactions of the American Math-
ematical Society 37(1), 1–20). This is available online at: http://www.ams.org/
journals/tran/1935-037-01/S0002-9947-1935-1501776-7/S0002-9947-1935-
1501776-7.pdf (accessed 10/10/2015). In this paper he addresses the idea of
Cauchy sequences in metric spaces and comments: “The need of uniformity in
[metric space] M arises from the fact that the elements of a fundamental sequence
are postulated to be ‘near to each other,’ and not near to any fixed point. As a gen-
eral topological space . . . has no property which leads itself to the definition of such
a ‘uniformity,’ it is impossible that a reasonable notion of ‘completeness’ could
be defined in it.” In this paper, von Neumann discusses total boundedness and
comapctness is the setting of topological linear spaces. His definition of complete is
43. Complete Metric Spaces 3

then:
Topological linear space L is topologically complete if every
closed and totally bounded set S ⊂ L is compact.
The ‘uniformity’ concern is dealt with by ‘anchoring’ open sets at the origin of the
linear space (that is, using the zero vector 0): “However, linear spaces . . . , even if
only topological, afford a possibility of ‘uniformization’ for their topology: because
of their homogeneity everything can be discussed in the neighborhood of 0.”

Note. For metric space (X, d) define the standard bounded metric d(x, y) =
min{d(x, y), 1}. Then a sequence (xn ) is Cauchy in (X, d) if and only if it is
Cauchy in (X, d) (since the Cauchy definition deals with “small” distances) and
is convergent in (X, d) if and only if it is convergent in (X, d). So space (X, d) is
complete if and only if (X, d) is complete.

Lemma 43.1. A metric space (X, d) is complete if every Cauchy sequence in X

has a convergent subsequence.

Example 1. The metric space Q with the usual metric is not complete, since
√
we can consider a sequence of rationals, (qn ), which converge to 2 (or any given
irrational) in R. Since the sequence converges in R then it is Cauchy in R and in Q.
But (qn ) does not converge in Q. Think of a Cauchy sequence as a sequence which
“wants” to to converge; in a complete space, it will converge. Cauchy sequences in
non-complete spaces may not converge, informally, because the space has a “hole”
at the point to which the sequence wants to converge!
43. Complete Metric Spaces 4

Theorem 43.2. Euclidean space Rk (where k ∈ N) is complete in either of its

usual metrics, the Euclidean metric d or the square metric ρ.

Note. We saw in Section 20 that with the standard bounded metric on R, d(a, b) =
min{|a − b|, 1}, we have the following as a metric on Rω :

D(x, y) = sup{d(xi, yi)/i}.

i∈N

By Theorem 20.5, D induces the product topology on Rω . We now show that Rω

is complete relative to D. First we need a lemma about convergence of sequences
in a product space.

Q
Lemma 43.3. Let X be the product space X = α∈J Xα (under the product
topology) and let (xn ) be a sequence of points in X. Then xn → x if and only if
πα (xn ) → πα (x) for all α ∈ J.

Theorem 43.4. There is a metric for the product space Rω relative to which Rω
is complete.

Note. Rω forms a linear space in the sense that for any a, b ∈ R and x, y ∈ Rω we
have ax + by ∈ Rω where we define ax + by = z with zi = axi + byi for all i ∈ N.
Since D is a metric on Rω , we can define a norm, k · k, on Rω as

kxk = D(x, 0) = sup{d(xi, 0)/i} = sup{min{|xi|, 1}/i} = sup{min{|xi|/i, 1/i}}.

i∈N i∈N i∈N

So Rω is a normed linear space (notice that kxk ≤ 1 for all x ∈ Rω ). By Theorem

43. Complete Metric Spaces 5

43.4, it is a complete normed linear space. A complete normed linear space is

called a Banach space. So Rω , along with k · k, is an example of a Banach space.
The “classical Banach spaces” are studied in our Real Analysis sequence (MATH
5210/5220) and based on Lebesgue integration theory. They are defined as
 Z  Z 1/p
Lp (E) = f  |f |p < ∞ where kf k = |f |p

E E

for 1 ≤ p < ∞. There is also an L∞ (E) space. The space L2 (E) is also a complete
inner product space (a complete inner product space is called a Hilbert space).

Note. In Example 21.1 it is shown that RJ where J is uncountable under the

product topology is not metrizable. We now introduce a metric on RJ relative to
which RJ is complete.

Definition. Let (Y, d) be a metric space. Let d(a, b) = min{d(a, b), 1} be the
standard bounded metric on Y derived from d. If x = (xα )α∈J and y = (yα)α∈J are
points in Y J then let

ρ(x, y) = sup{d(xα, yα ) | α ∈ J}.

ρ is the uniform metric on Y J corresponding to d.

Note. Recall that an element of Y J is a function f : J → Y , so that for f, g ∈ Y J

we have the notation

ρ(f, g) = sup{d(f (α), g(α)) | α ∈ J}.

43. Complete Metric Spaces 6

Theorem 43.5. If the space Y is complete in the metric d, then the space Y J is
complete in the uniform metric ρ corresponding to d.

Note. Since an element of Y J is a function f : X → Y , if X and Y are both

topological spaces then f ∈ Y X is a function f : X → Y and we can test function
f for continuity on space X. The set of all continuous functions from topological
space X to topological space Y is denoted C(X, Y ).

Definition. Let (Y, d) be a metric space and X a set. Function f : X → Y is

bounded if f (X) is a bounded subset of metric space (Y, d) (that is, there exists
M ∈ R such that for all y1, y2 ∈ f (X) we have d(y1 , y2) ≤ M ). The set of all
bounded functions from set X to metric space (Y, d) is denoted B(X, Y ).

Note. The next result shows that if X is a topological space and Y is a complete
metric space, then both C(X, Y ) and B(X, Y ) are complete in the uniform metric.

Theorem 43.6. Let X be a topological space and let (Y, d) be a metric space. The
set C(X, Y ) of continuous functions is closed in Y X under the uniform metric. So is
the set B(X, Y ) of bounded functions. Therefore, if Y is a complete metric space,
then both C(X, Y ) and B(X, Y ) are complete metric spaces under the uniform
metric.
43. Complete Metric Spaces 7

Note. If E ⊂ R then the set of all bounded functions mapping E into R forms a
“linear space.” A norm on this linear space if given by kf k∞ = sup{|f (x)| | x ∈ E}
and a metric is given by d(f, g) = kf − gk∞ = sup{|f (x) − g(x)| | x ∈ E}. You
would encounter the is in the setting of “Lp spaces” in Real Analysis (MATH
5210/5220). We take this as motivation for the following definition.

Definition. Let (Y, d) be a metric space. For f, g ∈ B(X, Y ) define

ρ(f, g) = sup{d(f (x), g(x)) | x ∈ X}.

This the the sup metric on B(X, Y ).

Note. The sup metric is related to the uniform metric. Let f, g ∈ B(X, Y ).
If ρ(f, g) > 1 then there is some x0 ∈ X such that d(f (x0 ), g(x0 )) > 1. Since
d(f (x), g(x)) = min{d(f (x), g(x)), 1} then d(f (x0 , g(x0 ) = 1. Since ρ(f, g) =
sup{d(f (x), g(x)) | x ∈ X} then, here, we ρ(f, g) = 1. On the other hand, if
ρ(f, g) ≤ 1 then d(f (x), g(x) = d(f (x), g(x)) ≤ 1 and so ρ(f, g) = ρ(f, g). In either
case, we have ρ(f, g) = min{ρ(f, g), 1}. This motivates the notation for ρ, as first
introduced in Section 20.

Note. The next result shows that every metric space can be embedded in a com-
plete metric space. An alternative proof is outlined in Exercise 43.9.
43. Complete Metric Spaces 8

Theorem 43.7. Let (X, d) be a metric space. There is an isometric embedding of

X into a complete space.

Definition. Let X be a metric space. If h : X → Y is an isometric embedding

of X into a complete metric space Y , then the subspace h(X) of Y is a complete
metric space (a closed subspace of a complete space contains all of its limit points
by the Sequence Lemma [Lemma 21.2], so every Cauchy sequence of elements of
k(X) converges to an element of h(X)) called the completion of X.

Note. In the previous definition, we use h(X) as the completion of X, as opposed

to complete space Y , in order to make the completion the “smallest” complete
space containing X (technically, containing the isometric image of X, h(X)). For
example, the completion of Q is R, although R2 is complete and contains Q. In
Exercise 43.10, it is shown that the completion of X is uniquely determined “up to
an isometry.”

Revised: 11/24/2018