Existence of Continuous Function
Existence of Continuous Function
Prabal Acharya
418MA2081
Department of Mathematics
National Institute of Technology Rourkela
Abstract
Urysohn’s lemma states that on a normal topological space disjoint closed subsets may be
seperated by continuous functions in the sense that a continuous function exists which takes
value 0 on one of the two subsets and value 1 on the other one (this type of function is called
an ”Urysohn function”). In fact the existence of such functions is equivalent to a space being
normal.
1 Introduction
If x 6= y are two distinct points of a space X, is there a continuous function f : X → R such that
f (x) 6= f (y)? In general this may not be true. There may not exist continuous functions on the
given space other than the constants. Let X be a space. Let τ = {φ, X} be a topology on X.
Then f : X → R is continuous if and only if f is constant. For each pair of distinct points, if there
is an f ∈ C(X, R) with f (x) 6= f (y), we say that the family C(X, R) separates points. This is the
reason for defining the completely regular and normal spaces which ensures plenty of continuous
functions. Indeed, the Urysohn’s lemma and its corollary, the Tietze extension theorem, provide
powerful tools for constructing continuous functions from x to R with some specified behaviour,
for any normal space X. They are among the most widely usued tools in topology.
1
2 Case of Metric spaces
The crucial fact here is the simple observation: If (X, d) is a metric space and x ∈ X, then the
function fx (y) := d(x, y) is continuous on X. For by triangle inequality we have,
Lemma 1 Let A be any nonempty subset of a metric space (X, d) and x ∈ X. Define d(x, A) ≡
dA (x) = inf {d(x, a) : a ∈ A}. Then |dA (x)−dA (y)| ≤ d(x, y) and hence dA is uniformly continuous
on X.
Theorem 1 dA (x) = 0 if and only if x is a limit point of A, where A is any nonempty subset of
a metric space (X, d) and dA (x) defined as above.
Proof : For any x ∈ X, d(x, A) + d(x, B) 6= 0. For, if it were so, then d(x, A) = 0 = d(x, B). Since
A and B are closed x ∈ A and x ∈ B by the last corollary. This contradicts our hypothesis that
A ∩ B = φ.
d(x, A)
The function f (x) = meets our requirments.
d(x, A) + d(x, B)
1 3
Example 1 ,
2 8
are dyadic rationals but not 13 .
2
Theorem 3 Dyadic Rationals are dense in R.
Proof: Let a and b be two real numbers such that a < b. By the Archimedean Property, there
exists n ∈ N such that
1 1 1
0< < b − a, which implies 0 < n < < b − a.
n 2 n
Thus we have 1 < b2n − a2n . As the distance between b2n and a2n is greater than 1, there exists
m ∈ N such that a2n < m < b2n which implies that a < 2mn < b, (2n 6= 0). So we proved that for
each open interval (a, b) ⊂ R, there exists a rational number of the form 2mn which belongs to (a, b).
In other words, the set of dyadic rationals is dense in R.
Lemma 3 A space X is normal space if and only if given a closed set F and an open set V
containing F there exists an open set U such that F ⊂ U ⊂ U ⊂ V.
Proof : Let X be normal and F, V as above. Then F and X\V are disjoint closed sets. By
normality of X there exist open sets U and W such that F ⊂ U and X\V ⊂ W and U ∩ W = φ.
Since U ⊂ X\W is closed, We see that Ū ⊂ X\W ⊂ V . Thus U is as required.
To prove the converse, suppose the condition on the existence of U holds, and let A, B be
two disjoint closed subsets of X. Then A ⊂ B c ⊂ X so there exists U open with A ⊂ U ⊂ Ū ⊂ B c .
But then U, Ū c are disjoint open sets with A ⊂ U and B ⊂ Ū c . So, X is normal.
Lemma 4 Let X be a normal space. If A and B are disjoint closed subsets of X, for each dyadic
rational r = k2−n ∈ (0, 1], there is an open set Ur with the following properties:
(i) A ⊂ Ur ⊂ X\B,
Proof: Let U1 = X\B. Since B is closed, U1 is open. Again A ⊂ U1 as A and B are disjoint.
Now by the last lemma, there exists disjoint open sets V and W such that
A ⊂ V ⊂ V̄ ⊂ U1 and B ⊂ W.
Let V = U 1 . Then
2
A ⊂ U 1 ⊂ Ū 1 ⊂ U1
2 2
3
Applying the same lemma once again to the open set U 21 containing A and to the open set U1
containing Ū 1 . We get open sets U 1 and U 3 such that
2 4 4
A ⊂ U 1 ⊂ Ū 1 ⊂ U 1 ⊂ Ū 1 ⊂ U 3 ⊂ Ū 3 ⊂ U1
4 4 2 2 4 4
Continuing this manner, we construct, for each dyadic rational r ∈ (0, 1), an open set Ur with the
following properties:
(ii) A ⊂ Ur , 0 < r ≤ 1.
(iii) Ur ⊂ U1 , 0 < r ≤ 1.
More formally, we proceed as follows. We select Ur for r = k2−n by induction on n. Assume that
we have chosen Ur for r = k2−n , 0 < k < 2n , 1 ≤ n ≤ N − 1. To find Ur for r = (2j + 1)2−N , 0 ≤
j < 2N −1 , observe that Ūj21−N and X\Ū(j+1)21−N are disjoint closed sets. So once again applying
to the last lemma, we can choose an open set Ur such that
Proof: Let r be adyadic rational in [0, 1]. Let Ur′ s be as in the last lemma. Now we define a
function (
0 if x ∈ Ur for all r,
f (x) =
sup{r : x ∈/ Ur } otherwise.
Clearly, 0 ≤ f ≤ 1. Let x ∈ A. Since A ⊂ Ur , 0 < r ≤ 1 ⇒ x ∈ Ur , ∀r. This implies f (x) = 0.
Since x is arbitrary f ≡ 0 on A.
Again let ∈ B. Then for any dyadic rational r, x ∈ / Ur . So, f (x) = sup{r : 0 < r ≤ 1} = 1. Since
x is arbitrary f ≡ 1 on B.
Now We need only establish the continuity of f .
Let x ∈ X be such that 0 < f (x) < 1. Let ǫ > 0. Choose dyadic rationals s and t in (0, 1) such
that f (x) − ǫ < s < f (x) < t < f (x) + ǫ. Then x ∈/ Up for dyadic rationals p ∈ (s, f (x)). By the
previous lemma x ∈ / Ūs . On other hand x ∈ Ut . Hence W = Ut \Ūs is an open neighbourhood of x.
Now we show that f (W ) ⊂ (f (x) − ǫ, f (x) + ǫ).
Let y ∈ W , then from the definition of f we see that s ≤ f (y) ≤ t. In particular |f (y) − f (x)| < ǫ
for y ∈ W . Thus f is continuous.
4
Corollary 4.1 Let X be a topological spaces in which all one-point sets are closed. Then X is
normal if and only of every pair of disjoint closed sets can be separated by a continuous function
in the sense that this continuous function takes value 0 on one of the two subsets and value 1 on
the other one( such function is called an ”Urysohn function”).
4 Applications
Urysohn’s lemma is a key in the proof of many other theorems, for instance:
• Urysohn Metrization Theorem: IfX is a normal space with a countable basis (i.e.
second countable), then we can use the abundance of continuous functions from X to [0, 1]
to assign numerical coordinates to the points of X an obtain an embedding of X into Rω .
From this we see that every second-countable normal space is a metric space.
5 Conclusion
Urysohn’s lemma is general result that holds in a large class of topologicals spaces. This lemma
is the surprising fact that being able to separate closed sets from one another with a continuous
function is not stronger than being able to separate them with open sets.
References
[1] Munkres.J, Topology, Pearson New International Edition, 2nd Edition.