3 COUNTABILITY AND CONNECTEDNESS AXIOMS

Definition 3.1 Let X be a topological space. A subset D of X is dense in X iff D̄ = X.

X is separable iff it contains a countable dense subset.
X satisfies the first axiom of countability or is first countable iff for each x ∈ X, there is a
countable neighbourhood basis at x. Any metrisable space is first countable.
X satisfies the second axiom of countability or is second countable iff the topology of X has
a countable basis.
For example, R is separable, Q forming a countable dense subset. R is even second countable,
for example the countable collection of open intervals with rational end-points forms a basis.
Lemma 3.2 Let D be a subset of a space X and let B be a basis of open sets not containing
∅. Then D is dense in X iff for each B ∈ B, B ∩ D 6= ∅.
Proof. ⇒: Suppose that B ∈ B but B ∩ D = ∅. Then D is contained in the closed set X − B,
so D̄ 6= X.
⇐: Let x ∈ X and let N be a neighbourhood of x in X. Then there is B ∈ B so that
x ∈ B ⊂ N . Since B ∩ D 6= ∅, it follows that N ∩ D 6= ∅, so by Proposition 1.17 x ∈ D̄, so
that D̄ = X.
Theorem 3.3 Let X be second countable. Then X is separable and first countable.
Proof. Let B be a countable basis: we may assume that ∅ ∈ / B.
For each B ∈ B, choose xB ∈ B. Then by Lemma 3.2, {xB / B ∈ B} is a countable dense
subset of X, so X is separable.
For each x ∈ X, {B ∈ B / x ∈ B} is a countable neighbourhood basis at x, so X is first
countable.
In general no other relations hold between these three properties. For example, an un-
countable discrete space is first countable but not separable and an uncountable space with the
cofinite topology is separable but not first countable. The real line with the right half-open
interval topology is separable and first countable but not second countable.
Theorem 3.4 The topological product of a countable family of separable (first countable, second
countable) spaces is separable (first countable, second countable).
Proof. Let {Xn / n = 1, 2, . . .} be a countable family of topological spaces and let X = ΠXn .
(i) Suppose that for each n, Xn is separable, say {xi,n / i = 1, 2, . . .} is a countable dense
subset. Let Dm = {hxi,n i ∈ X / i = 1 whenever n > m}. Then Dm is countable, and so
also is D = ∪Dm . If U = ΠUn is a basic open set with U 6= ∅, then there is m so that
for each n > m, Un = Xn . For each n ≤ m, there is xi,n ∈ Un . Thus U ∩ Dm 6= ∅, so
U ∩ D 6= ∅ and hence by Lemma 3.2, D is dense and so X is separable.
(ii) Suppose that for each n, Xn is first countable. Let x ∈ X, and for each n, let Mn be a
countable neighbourhood basis at xn . We may assume that each member of Mn is open
in Xn . Consider
M = {ΠAn / An = Xn for all but finitely many n; if An 6= Xn then An ∈ Mn }.
Then M is a countable family of open neighbourhoods of x. Further, if N is any neigh-
bourhood of x, then x ∈ ΠUn ⊂ N for some Un open in Xn with Un 6= Xn for only finitely
many n. If Un = Xn , let An = Xn , and if Un 6= Xn , then there is An ∈ Mn so that
x ∈ An ⊂ Un . Then ΠAn ∈ M and x ∈ ΠAn ⊂ ΠUn ⊂ N . Thus M is a countable
neighbourhood basis at x, so X is first countable.

21
(iii) Suppose that for each n, Xn is second countable, say Bn is a countable basis. Then

{ΠBn / Bn = Xn for all but finitely many n; if Bn 6= Xn then Bn ∈ Bn }

is a countable basis for X, so X is second countable.

As might be expected, we cannot replace the countable product in Theorem 3.4 by an

arbitrary product. For example, let A be an uncountable index set and for each α ∈ A, let Xα
be the set of positive integers with the discrete topology. Then {{x} / x ∈ Xα } is a countable
basis for Xα , so Xα is second countable, and hence separable and first countable.
The space X = ΠXα is not first countable, for if {Ni } is a countable family of neighbour-
hoods of x ∈ X, then for each i, πα (Ni ) = Xα for all but finitely many indices α. Thus there
is β ∈ A so that for each i, πβ (Ni ) = Xβ . Then πβ−1 (xβ ) is a neighbourhood of x containing no
Ni , so {Ni } cannot be a neighbourhood basis at x.
Suppose that D is a dense subset of X. Let P(D) denote the power set of D and define
ϕ : A → P(D) by ϕ(α) = D ∩ πα−1 (1). If α, β ∈ A with α 6= β, then

ϕ(α) − ϕ(β) = D ∩ [πα−1 (1) ∩ πβ−1 (Xβ − {1})] 6= ∅,

so ϕ is injective. Thus card A ≤ card P(D). Hence if card A > 2ℵ0 , then card D > ℵ0 , i.e. D
is uncountable: thus in this case X is not separable.

Definition 3.5 Let X be a space. Say that a sequence hxn i in X converges to x ∈ X and write
limn→∞ xn = x iff for each neighbourhood N of x, there is n0 so that for each n ≥ n0 , xn ∈ N .
If there is n0 so that for each n ≥ n0 , xn ∈ N , say that hxn i is eventually in N .

Convergent sequences are not of as much use in topological spaces as they are in metric
spaces. There are two particular problems. Firstly, in a general topological space the limit need
not be unique: in fact in an indiscrete space, every sequence converges to every point. Secondly,
convergent sequences do not capture the topology near a point: this disadvantage is highlighted
in Example 3.9. Nets and filters have been evolved to overcome the main problems but we do
not consider either.

Proposition 3.6 Let X be a Hausdorff space, and let hxn i be a sequence in X. If limn→∞ xn =
x and limn→∞ xn = y then x = y.

Proof. If x 6= y then there are neighbourhoods U of x and V of y so that U ∩ V = ∅. It is

impossible for any sequence to be eventually in both U and V .

Theorem 3.7 Let X be a first countable space and let A ⊂ X. Then

Ā = {x ∈ X / there is a sequence hxn i in A with lim xn = x}.

n→∞

Proof. If hxn i is a sequence in A and limn→∞ xn = x, then x ∈ Ā by Proposition 1.17 and the
definition of limn→∞ xn = x.
Conversely, suppose that x ∈ Ā. Let {Ni / i = 1, 2, . . .} be a countable neighbourhood basis
at x. We may assume that for each i, Ni+1 ⊂ Ni ; otherwise replace each Ni by ∩ij=1 Nj . Now
for each i, A ∩ Ni 6= ∅; choose xi ∈ A ∩ Ni . Then hxn i is a sequence in A and limn→∞ xn = x.

Corollary 3.8 Let X be a first countable space and let f : X → Y be a function. Then f is
continuous at x ∈ X iff for every sequence hxn i of points of X such that limn→∞ xn = x, we
have limn→∞ f (xn ) = f (x).

22
Proof. ⇒: obvious.
⇐: we use the criterion of problem 10 of section 1. Let A ⊂ X be any set with x ∈ Ā.
By Theorem 3.7, there is a sequence hxn i in A for which limn→∞ xn = x. By hypothesis,
limn→∞ f (xn ) = f (x). Since f (xn ) ∈ f (A), we have f (x) ∈ f (A) as required, so f is continuous.

Example 3.9 The criteria in Theorem 3.7 and Corollary 3.8 are invalid if we remove the
hypothesis of first countability.

For example, let X be any uncountable set and topologise X by the cocountable topology
of problem 3 of section 1. Any convergent sequence in X is eventually constant but if A is
any proper uncountable subset of X then Ā 6= A so Theorem 3.7 is invalid if we delete first
countability. Let Y have the same underlying set as X but with the discrete topology and let
f : X → Y be given by f (x) = x. Then f is not continuous although if hxn i is a sequence of
points of X and limn→∞ xn = x then limn→∞ f (xn ) = f (x).

Theorem 3.10 Every regular second countable space is normal.

Proof. Let X be a regular second countable space, say B is a countable basis, and suppose that
A and B are disjoint closed subsets of X.
For each x ∈ A, the set X − B is a neighbourhood of x, so by regularity of X, there is
Ux ∈ B so that x ∈ Ux ⊂ Ux ⊂ X − B. Now {Ux / x ∈ A} is countable, so we may reindex and
rename to {Ui / i = 1, 2, . . .}. Then the countable collection {Ui / i = 1, 2, . . .} of open sets
satisfies: A ⊂ ∪i Ui ; and for each i, B ∩ Ui = ∅.
Similarly we can find a countable collection {Vi / i = 1, 2, . . .} of open sets satisfying:
B ⊂ ∪i Vi ; and for each i, A ∩ Vi = ∅.
Define the sets Yi and Zi (i = 1, 2, . . .) by Yi = Ui − [∪in=1 Vn ] and Zi = Vi − [∪in=1 Un ]. Then
Yi and Zi are also open sets. Let U = ∪i Yi and V = ∪i Zi , both open sets. It is claimed that U
and V satisfy the requirements of the definition of normality.
U ∩V = ∅, for if x ∈ U ∩V , then there are i, j so that x ∈ Yi ∩Zj . We might as well suppose
that i ≥ j, in which case x ∈ Yi = Ui − [∪in=1 Vn ] ⊂ Ui − Vj and x ∈ Zj ⊂ Vj , a contradiction.
Thus U ∩ V = ∅.
A ⊂ U , since A ⊂ ∪i Ui and A ∩ Vj = ∅. Similarly B ⊂ V .
Thus U and V are disjoint open neighbourhoods of A and B, so X is normal.

Theorem 3.11 (Urysohn’s metrisation theorem) Every second countable T3 space is metris-
able.

Proof. Let X be a second countable T3 space, and let B be a countable basis for X. We
construct a metric on X in two steps.

(i) There is a countable family {gn : X → [0, 1] / n = 1, 2, . . .} of continuous functions

satisfying: for each x ∈ X and each neighbourhood N of x, there is n so that gn (x) > 0
and gn (X − N ) = 0. Indeed by Theorems 3.10 and 2.20, for each pair U, V ∈ B for
which Ū ⊂ V , there is a continuous function gU,V : X → [0, 1] so that gU,V (Ū ) = 1 and
gU,V (X −V ) = 0. The collection {gU,V / U, V ∈ B and Ū ⊂ V } is countable. Furthermore,
if x ∈ X and N is a neighbourhood of x then there are U, V ∈ B so that x ∈ U ⊂ Ū ⊂
V ⊂ N : for such U and V , we have gU,V (x) > 0 and gU,V (X − N ) = 0. Now relabel the
countable family {gU,V / U, V ∈ B and Ū ⊂ V } as {gn : X → [0, 1] / n = 1, 2, . . .}.

23
 (ii) Given the family {gn : X → [0, 1] / n = 1, 2, . . .} of step (i), define fn : X → [0, 1] by
fn (x) = gnn(x) . Now define d : X × X → [0, 1] by

d(x, y) = lub{|fn (x) − fn (y)| / n = 1, 2, . . .}.

It remains to verify that (A) d is a metric; and (B) the topology induced by d is the topology
on X.

(A) d is a metric.

(a) d(x, x) = 0
(b) If d(x, y) = 0 then x = y, for if x 6= y, then X − {y} is a neighbourhood of x so there
is n for which fn (x) > 0 and fn (y) = 0. Thus d(x, y) ≥ |fn (x) − fn (y)| > 0.
(c) d(x, y) = d(y, x)
(d) d(x, z) ≤ d(x, y) + d(y, z), because for each n,

|fn (x) − fn (z)| ≤ |fn (x) − fn (y)| + |fn (y) − fn (z)| ≤ d(x, y) + d(y, z),

so d(x, y) + d(y, z) is an upper bound of the set for which d(x, z) is the least upper
bound.

(B) the topology induced by d is the topology on X.

(a) let x ∈ X and let N be a neighbourhood of x. Choose n for which fn (x) > 0
and fn (X − N ) = 0, and let ε = fn (x). Then for any y ∈ X, if d(x, y) < ε then
|fn (x) − fn (y)| < ε so fn (y) > 0 and hence y ∈ N .
Thus N is also a neighbourhood of x in (X, d).
(b) let x ∈ X and let ε > 0. Let m = 1ε .
 

For each n ≤ m, let Nn = fn−1 ((fn (x) − ε, fn (x) + ε)). Then for each y ∈ Nn ,
|fn (x) − fn (y)| < ε. Also, for each n > m we have n1 < ε, so that for each y ∈ X and
each n > m, |fn (x) − fn (y)| ≤ n1 < ε.
Set N = ∩m n=1 Nn . Then N is a neighbourhood of x since each Nn is; and if y ∈ N ,
then
 
1
d(x, y) ≤ max , |f1 (x) − f1 (y)|, |f2 (x) − f2 (y)|, . . . , |fm (x) − fm (y)| < ε,
m+1
so N is contained in the d-ball of radius ε centred at x.

By (a) and (b), the neighbourhood systems for the two topologies coincide. Thus by
Proposition 1.11, the two topologies themselves are the same.

Let 2 denote the discrete space whose underlying set is {0, 1}. This space is the prototypical
disconnected space.

Definition 3.12 A space X is connected iff every continuous function f : X → 2 is constant.

Otherwise the space is disconnected, and if X is disconnected then we will call a continuous
surjection δ : X → 2 a disconnection (of X).
A subset C of a space X is connected iff the subspace C of X is connected.
Say that two points x, y ∈ X are connected in X iff there is a connected subset C of X
containing both x and y. The relation “are connected in” is an equivalence relation. Call the
equivalence class of x ∈ X under this relation the component of x in X.

24
 Many proofs involving connectedness make use of a disconnection and argument by reductio
ad absurdum. Connectedness and disconnectedness are both obviously topological properties.

Theorem 3.13 Let X be a space. The following are equivalent:

(i) X is connected;

(ii) every two points of X are connected in X;

(iii) X cannot be expressed as the union of two non-empty disjoint open subsets;

(iv) X cannot be expressed as the union of two non-empty disjoint closed subsets;

(v) the only subsets of X which are both open and closed are ∅ and X.

Proof. (i)⇒(ii): trivial.

(ii)⇒(iii): suppose that X = U ∪ V , where U ∩ V = ∅ and U and V are non-empty open
sets. Pick x ∈ U and y ∈ V . Let A ⊂ X be any set containing x and y. Then δ : A → 2 defined
by δ(A ∩ U ) = 0 and δ(A ∩ V ) = 1 is a disconnection of A, so x and y are not connected in X.
(iii)⇒(iv): if X = C ∪ D, where C ∩ D = ∅ and C and D are non-empty closed sets, then
X = (X − C) ∪ (X − D), a union of two non-empty disjoint open sets.
(iv)⇒(v): if S is both open and closed in X but ∅ 6= S 6= X, then S and X − S are two
non-empty disjoint closed sets whose union is X.
(v)⇒(i): if δ : X → 2 is a disconnection of X, then δ −1 (0) is both open and closed but is
neither ∅ nor X.

Theorem 3.14 Let {Cα / α ∈ A} be a family of connected subsets of a space X satisfying:

there is β ∈ A so that for each α either Cα ∩ Cβ 6= ∅ or Cα ∩ Cβ 6= ∅. Then C = ∪α∈A Cα is
connected.

Proof. Consider firstly the case where C = X. Suppose instead that X is not connected. Then
by Theorem 3.13, we may write X = U ∪V , where U and V are non-empty disjoint open subsets
of X. Since Cβ is connected, it must lie inside one of U and V , say Cβ ⊂ U . Then for each α,
we must have Cα ⊂ U for by connectedness, either Cα ⊂ U or Cα ⊂ V , but if α ∈ A is such
that Cα ⊂ V , then Cβ ⊂ Ū ⊂ X − V , so Cα ∩ Cβ = ∅ and Cα ⊂ V̄ ⊂ X − U , so Cα ∩ Cβ = ∅,
contrary to hypothesis. Thus for each α ∈ A, Cα ⊂ U , so X ⊂ U , and V must be empty, a
contradiction. Thus X is connected.
If C 6= X then any point of Cα ∩ Cβ must be in C, from which it follows that either
Cα ∩ Cβ 6= ∅ or Cα ∩ Cβ 6= ∅ (where closure here refers to closure in C), and hence we may
apply the previous case to the subspace C.
Note that a family {Cα / α ∈ A} of connected subsets having non-empty intersection satisfies
the hypotheses of the theorem.

Corollary 3.15 Suppose that C is a connected subset of X and A ⊂ X satisfies C ⊂ A ⊂ C̄.

Then A is connected.

Proof. Apply Theorem 3.14 to the family {C} ∪ {{x} / x ∈ A}.

Theorem 3.16 For each x in a topological space X, the component of x in X is the largest
connected subset of X containing x. Moreover, the component is closed.

25
Proof. Let C denote the component of x in X. Clearly if K is any connected subset of X
containing x then K ⊂ C. On the other hand, for each y ∈ C, there is a connected set Ky ⊂ X
containing both x and y. By Theorem 3.14, ∪y∈C Ky is connected. Thus C = ∪y∈C Ky is
connected, and hence is the largest connected set containing x.
By Corollary 3.15, C̄ is connected, so C̄ = C, i.e. C is closed.
Components need not be open. For example, giving the rationals, Q, the usual topology
inherited from R, the only connected sets are ∅ and the singletons, so the only components are
the singletons, none of which is open.

Theorem 3.17 The topological product of connected spaces is connected.

Proof. The proof proceeds in three stages, considering the product of two spaces, then of finitely
many spaces, and finally a product of arbitrarily many spaces.
Let X and Y be connected spaces and let (x1 , y1 ) and (x2 , y2 ) be any two points of X × Y .
Since {x1 } × Y is homeomorphic to Y , it is connected. X × {y2 }, being homeomorphic to
X, is connected. Moreover, (x1 , y2 ) ∈ ({x1 } × Y ) ∩ (X × {y2 }). Thus by Theorem 3.14,
({x1 } × Y ) ∩ (X × {y2 }) is connected: this set contains both (x1 , y1 ) and (x2 , y2 ), so by Theorem
3.13(ii), X × Y is connected.
By induction, a finite product of connected spaces is connected.
Suppose now that {Xα / α ∈ A} is any family of connected spaces. Let X = ΠXα , let
x ∈ X, and let C be the component of x in X. It is sufficient to show that if U is any non-
empty basic open set then U ∩ C 6= ∅, for then by Lemma 3.2, C̄ = X. However, Theorem 3.16
tells us that C is closed. Thus C = X, i.e., X is connected.
Suppose, then, that U = ΠUα 6= ∅, where each Uα is open in Xα and Uα 6= Xα for only
finitely many indices α, say α1 , . . . , αn . For each i = 1, . . . , n, pick ui ∈ Uαi . Let y ∈ X
be the point defined by yα = xα for α ∈ / {α1 , . . . , αn }, and yαi = ui . Then y ∈ U . Now
K = {z ∈ X / zα = xα for each α ∈ A − {α1 , . . . , αn }} is homeomorphic to Xα1 × . . . × Xαn
so is connected by the finite case of the theorem already proven. Further, x ∈ K, so K ⊂ C.
Since y ∈ K, we have y ∈ U ∩ C, so U ∩ C 6= ∅.

Proposition 3.18 Let f : X → Y be continuous and let C be a connected subset of X. Then

f (C) is connected.

Proof. If not, then there is a disconnection δ : f (C) → 2. Then δf : C → 2 is also a

disconnection.

Theorem 3.19 A subset of R (usual topology) is connected iff it is an interval.

Proof. ⇒: Let A ⊂ R. If A is not an interval, then there is c ∈ R − A and a, b ∈ A so that

a < c < b. Define δ : A → 2 by δ(x) = 0 for all x < c and δ(x) = 1 for all x > c . Then δ is a
disconnection of A.
⇐: Let A ⊂ R be an interval. If δ : A → 2 is a disconnection of A, then let a ∈ δ −1 (0) and
b ∈ δ −1 (1). We may suppose that a < b. The non-empty set B = {x ∈ A / x < b and δ(x) = 0}
A A
is bounded above by b; let α = lubB. By Proposition 1.17, α ∈ B , where B denotes the
closure of B in A. What is δ(α)?
A
By Theorem 1.19(vi), δ(B ) ⊂ {0} = {0}, so δ(α) = 0. Since δ(α) = 0 but δ(b) = 1 we
A
have (α, b] 6= ∅, so δ((α, b]) = {1}, and by Theorem 1.19(vi), δ([α, b]) = δ((α, b] ) = {1} = {1},
so δ(α) = 1, a contradiction.

26
Theorem 3.20 (Intermediate Value Theorem) Let f : X → R be continuous, where X is
connected. Suppose x1 , x2 ∈ X and y ∈ R are such that f (x1 ) ≤ y ≤ f (x2 ). Then there is
x ∈ X such that f (x) = y.

Proof. By Theorem 3.19, f (X) is an interval, so y ∈ f (X).

Corollary 3.21 Let a, b ∈ R with a < b. Give [a, b] the usual topology, and let f : [a, b] → [a, b]
be continuous. Then there is a point x ∈ [a, b] such that f (x) = x.

Proof. Define g : [a, b] → R by g(x) = x − f (x). Then g is continuous, g(a) = a − f (a) ≤ 0,

and g(b) = b − f (b) ≥ 0. Thus by Theorem 3.20, there is x ∈ [a, b] such that g(x) = 0, i.e.,
f (x) = x.
Corollary 3.21 is known as Brouwer’s fixed-point theorem in dimension 1. A point x ∈ X is a
fixed-point of a function f : X → X iff f (x) = x. A space X has the fixed-point property iff every
continuous function f : X → X has a fixed-point. The fixed-point property is a topological
property. Brouwer’s fixed-point theorem in general says that B n = {x ∈ Rn / |x| ≤ 1} has the
fixed-point property; for a proof of this for n = 2 see chapter 5.
Using algebraic topology, one can assign to continuous functions on a certain class of topo-
logical spaces a number known as the Lefschetz number. The Lefschetz fixed-point theorem
says that if this number is non-zero then the function has a fixed-point. It turns out that when
the domain and range of the function are B n then the Lefschetz number is always non-zero.
The solution of ordinary differential equations involves finding a fixed-point of a certain
function. For example, if we wish to Rsolve y 0 = F (x, y) subject to y = y0 when x = x0 , then we
x
seek to solve the equation y = y0 + x0 F (x, y)dx. Let X denote the space of all continuously
Rx
differentiable functions from R to R and define L : X → X by L(f )(x) = y0 + x0 F (x, f (x))dx.
By suitably topologising X, L is continuous (provided F is nice enough!). A solution of y 0 =
F (x, y) is a fixed-point of L.

Proposition 3.22 Every connected Tychonoff space having at least two points must have at
least 2ℵ0 points.

Proof. Let X be a connected Tychonoff space having at least two points. It suffices to find a
surjection X → [0, 1]. Let x1 , x2 ∈ X be such that x1 6= x2 . Then {x2 } is closed, so there is
a continuous function f : X → [0, 1] with f (x1 ) = 0 and f (x2 ) = 1. By Theorem 3.20, f is a
surjection.

Definition 3.23 Let X be a space. A path in X is a continuous function π : [0, 1] → X, where

[0, 1] has the usual topology. If π(0) = x and π(1) = y, then the path is from x to y.
Two points x, y ∈ X are path connected in X iff there is a path in X from x to y.
X is path connected iff every pair of points of X are path connected in X.
As in Definition 3.12, the relation “are path connected in X” is an equivalence relation, so
gives rise to the idea of the path component of a point in a space, i.e. the largest path connected
subset containing that point.

Sometimes it is convenient to replace [0, 1] by some other closed interval. By Proposition

3.18 and Theorem 3.19, if two points are path connected in X then they are connected in X.
Clearly path connected spaces are connected.

Example 3.24 (i) connected spaces need not be path connected;

(ii) path components need not be closed;

27
(iii) the path analogues of Theorem 3.14 and Corollary 3.15 are not valid.

Consider the two subsets

 
2 1
X1 = (x1 , x2 ) ∈ R / 0 < x1 < ∞, x2 = sin and X2 = {(0, x2 ) ∈ R2 / − 1 ≤ x2 ≤ 1}
x1
of R2 . Let X = X1 ∪ X2 and give X the usual topology. Note that in X, X1 = X. Since
X1 is homeomorphic to (0, ∞) by projection on the first coordinate, by Theorem 3.19, X1 is
connected, so by Corollary 3.15, X is connected. However, it can be shown that X is not path
connected. The path components of X are X1 and X2 .

Theorem 3.25 Let f : X → Y be continuous and C a path-connected subset of X. Then f (C)

is path-connected.

Proof. If y1 , y2 ∈ f (C) then there are x1 , x2 ∈ C so that f (xi ) = yi . Let π : [0, 1] → C be a

path from x1 to x2 . Then f π is a path from y1 to y2 .

Theorem 3.26 The topological product of path-connected spaces is path-connected.

Proof. Let {Xα / α ∈ A} be a family of path-connected spaces, let X = ΠXα and let x, y ∈ ΠXα .
Then for each α ∈ A, there is a path pα : [0, 1] → Xα such that pα (0) = xα and pα (1) = yα .
Define p : [0, 1] → X by p(t)α = pα (t). By Proposition 1.26 p is a path from p(0) = x to
p(1) = y.

Definition 3.27 A space X is locally connected at x ∈ X iff for each neighbourhood U of x,

there is a neighbourhood V of x so that every two points of V are connected in U .
X is locally path-connected at x ∈ X iff for each neighbourhood U of x, there is a neigh-
bourhood V of x so that every two points of V are path-connected in U .
Say that X is locally (path-)connected iff X is locally (path-)connected at each x ∈ X.

Lemma 3.28 A space X is locally (path-)connected at x ∈ X iff every neighbourhood of x

contains a (path-)connected neighbourhood of x.

Proof. ⇒: Let U be a neighbourhood of x. Then there is a neighbourhood V of x so that

for each y ∈ V , x and y are (path-)connected in U . For each y ∈ V , let Cy be a connected set
containing x and y (a path from x to y) so that Cy ⊂ U . Let N = ∪y∈V Cy . Then V ⊂ N , so
N is a neighbourhood of x. Also N ⊂ U and N is (path-)connected.
⇐: trivial.

Theorem 3.29 Let X be a space. The following are equivalent:

(i) X is locally (path-)connected;
(ii) the (path) components of every open subspace of X are open in X;
(iii) the (path-)connected open subsets of X form a basis for the topology of X.

Proof. (i)⇒(ii): Let U be an open subspace of X. Clearly U is locally (path-) connected by

(i), so by Lemma 3.28 the (path) components of U are open in U and hence in X.
(ii)⇒(iii): Let U be any open subspace of X. By (ii), the (path) components of U are open
in X. Hence U is the union of a collection of (path-) connected open subsets of X, so such sets
form a basis.
(iii)⇒(i): Suppose x ∈ X and U is an open neighbourhood of x. By (iii), U is the union
of a collection of (path-)connected open sets. One of these must contain x, so X is locally
(path-)connected.

28
Theorem 3.30 Let X be a locally path-connected space. Then each path component of X is
both open and closed in X.

Proof. Let C be any path component of X. Then by Theorem 3.29(ii), C is open, hence all
path components are open. X − C, the union of all path components other than C, is open, so
C is closed.

Corollary 3.31 Every connected, locally path-connected space is path-connected.

Proof. By Theorems 3.13(v) and 3.30.

Exercises

1. Let X be a second countable space. Show that every family of disjoint open subsets of X
is countable.

2. Let (X, T ) be a second countable topological space and let B be any basis for T . Prove
that B contains a countable subfamily which is a basis for T .

3. Let X be a metrisable, separable space. Prove that X is second countable. Can we replace
“metrisable” by “first countable” in this statement?

4. Prove that the right half-open interval topology of Example 1.4 is not second countable
but is separable. Deduce from Exercise 3 that this topology is not metrisable.
P∞ 2
5. Let l2 = hx1 , x2 , . . .i / xi ∈ R and 2 2

i=1 xi < ∞ . Show that d : l ×l → [0, ∞) defined
P∞ 1
2 2 , where x = hx , x , . . .i, y = hy , y , . . .i ∈ l2 , is a metric

by d(x, y) = i=1 (xi − yi ) 1 2 1 2
on l2 . Prove that l2 with this metric is separable.

6. Let C be a subset of a topological space. Prove the equivalence of the following three
conditions:

(i) C is disconnected;
(ii) there are C1 , C2 ⊂ X such that C = C1 ∪ C2 , C1 ∩ C2 = ∅ = C1 ∩ C2 , C1 6= ∅,
C2 6= ∅.
(iii) there are open sets U, V ⊂ X such that C ⊂ U ∪ V , C ∩ U ∩ V = ∅, C ∩ U 6= ∅,
C ∩ V 6= ∅.

7. Consider the possibility of replacing C ∩ U ∩ V = ∅ in Exercise 6(iii) by U ∩ V = ∅.

(a) Verify that such a replacement is valid if X is metrisable.

(b) Give a counterexample illustrating that such a replacement is invalid in a general
topological space.

8. Consider the two statements about a topological space X:

(a) X is connected;
(b) for each subset Y of X for which ∅ 6= Y 6= X, frY 6= ∅.

Decide whether each of the implications (a)⇒(b) and (b)⇒(a) is true or false, giving a
complete justification for your answer.

29
 9. The Customs Passage Theorem asserts that for one to bring goods into a country from
outside, one must cross the frontier (so customs officials need only guard the frontier).
Put this theorem into a topologically precise form using the concepts of this chapter and
prove your version of the theorem.

10. Let Rω denote the product of a countable infinity of copies of R (i.e. Rω = Πn∈ω Xn ,
where ω = {0, 1, 2, . . .} and Xn = R for each n). Show that with the box topology, Rω is
not connected. (Hint: consider the subset consisting of all bounded sequences).

11. Prove that a space X is connected iff for each open cover {Uα / α ∈ A} of X and for
each x1 , x2 ∈ X, there is a finite sequence α1 , . . . , αn ∈ A so that x1 ∈ Uα1 , x2 ∈ Uαn and
Uαi ∩ Uαi+1 6= ∅ for each i = 1, . . . , n − 1.

12. Show that R and Rn (n > 1) are not homeomorphic.

13. Let X be as in Exercise 1.18. Prove that X is separable but not second countable. Deduce
that X is not metrisable.

14. Prove that every separable, locally connected space has only countably many components.
Give an example of a separable space having uncountably many components.

30