Topology (H) Lecture 3
Lecturer: Zuoqin Wang
Time: March 15, 2021
TOPOLOGY: DEFINITIONS AND EXAMPLES
1. Continuous maps between metric spaces: continued
¶ More metrics that induce equivalent conceptions of continuity.
Let’s study one more example.
Example 1.1. Consider another pair of metrics on Rn , say the Euclidian metric d2 (x, y) =
|x − y| and the bounded metric
d2 (x, y)
d¯2 (x, y) :=
1 + d2 (x, y)
induced by d2 . Obviously
d¯2 (x, y) ≤ d2 (x, y).
But d2 and d¯2 are not strongly equivalent, since given any constant C > 0, there
exists x, y ∈ Rn such that
d2 (x, y) > C ≥ C d¯2 (x, y).
However, if we study the conception of continuity of a function f : Rn → R, again we
will arrive at the same conclusion: a function f : (Rn , d2 ) → R is continuous if and
only if the function f : (Rn , d¯2 ) → R is continuous:
In fact, if f : (Rn , d¯2 ) → R is continuous, then f : (Rn , d2 ) → R is
continuous since d¯2 (x, y) ≤ d2 (x, y).
Conversely if f : (Rn , d2 ) → R is continuous, namely,
∀ε > 0, ∃δ > 0 s.t. ∀x ∈ X, d2 (x, x0 ) < δ =⇒ |f (x) − f (x0 )| < ε.
Since
δ d2 (x, y) δ
d¯2 (x, y) < =⇒ < =⇒ d2 (x, y) < δ,
1+δ 1 + d2 (x, y) 1+δ
we find that
δ
∀ε > 0, ∃δ 0 = > 0 s.t. ∀x ∈ X, d¯2 (x, x0 ) < δ 0 =⇒ |f (x) − f (x0 )| < ε.
1+δ
In other words, f : (Rn , d¯2 ) → R is continuous.
From the above examples one can imagine that there should be some underlying
structure that is more fundamental than the metric structure that induces the concep-
tion of continuity.
1
2 TOPOLOGY: DEFINITIONS AND EXAMPLES
¶ Local continuity via neighborhoods.
To figure out the structure behind continuity, let’s recall
A map f : (X, dX ) → (Y, dY ) is continuous at x0 ∈ X
⇐⇒∀ε > 0, ∃δ > 0 s.t. ∀x ∈ X, dX (x, x0 ) < δ, we have dY (f (x), f (x0 )) < ε
⇐⇒∀ε > 0, ∃δ > 0 s.t. B X (x0 , δ) ⊂ f −1 B Y (f (x0 ), ε) .
Of course these equivalent characterizations all depends on the metric structure
(either in terms of the metric d, or in terms of the metric balls). To get rid of the
“metric dependence”, let’s recall that a subset U ⊂ X is open if
∀x ∈ U, ∃ε = ε(x) > 0 such that B(x, ε) ⊂ U.
Intuitively, continuity of f at a point x0 concerns only points in X near x0 and points
in Y near f (x). Using open sets, we can introduce the following definition of neighbor-
hoods in which the metric does not appear explicitly:
Definition 1.2. We say a subset N ⊂ X is a neighborhood of x if there exists an open
set U in X so that x ∈ U ⊂ N .1
Remark 1.3. If we denote by N (x) the set of all neighborhoods of x, it is easy to see
(1) If N ∈ N (x), then x ∈ N.
(2) If M ⊃ N and N ∈ N (x) then M ∈ N (x).
(3) If N1 , N2 ∈ N (x), then N1 ∩ N2 ∈ N (x).
(4) If N ∈ N (x), then ∃M ⊂ N and M ∈ N (x), s.t. ∀y ∈ M, N ∈ N (y).
It turns out that we can characterize continuity at a point via neighborhoods:
Proposition 1.4. Let f : (X, dX ) → (Y, dY ) be a map between metric spaces. Then f
is continuous at x ∈ X if and only if the pre-image of any neighborhood of f (x) is a
neighborhood of x.
Proof. Suppose f is continuous at x ∈ X, and M ⊂ Y is a neighborhood of x. Then
by definition, there is an open set V ⊂ Y such that f (x) ∈ V ⊂ M . By definition of
open set, ∃ε > 0 s.t. B(f (x), ε) ⊂ V. By continuity of f at x, ∃δ > 0 such that
B(x, δ) ⊂ f −1 (B(f (x), ε)) ⊂ f −1 (V ) ⊂ f −1 (M ).
So f −1 (M ) is a neighborhood of x.
Conversely suppose for any neighborhood M ⊂ Y of f (x), f −1 (M ) is a neighbor-
hood of x. Then in particular for ∀ε > 0, f −1 (B(f (x), ε)) is a neighborhood of x,
i.e. it contains an open set U with x ∈ U. By the definition of open set, ∃δ > 0 s.t.
B(x, δ) ⊂ U, which implies B(x, δ) ⊂ f −1 (B(f (x), ε)). So f is continuous at x.
1In
some books (including Munkres’ book) people require neighborhoods to be open. We will not
make such requirement. Instead, we use the expression “an open neighborhood of x” to indicate a set
which is both open and is a neighborhood of x.
TOPOLOGY: DEFINITIONS AND EXAMPLES 3
Remark 1.5. In general, if f is continuous at x0 , it may happen that the pre-image of
on open neighborhood of f (x0 ) is not open in X. [Try to find an example!]
¶ Global continuity via open sets.
As a consequence of Proposition 1.4, we get the following characterization of (glob-
ally) continuous maps between abstract metric spaces:
Theorem 1.6. A map f : (X, dX ) → (Y, dY ) is continuous if and only if for any open
set V in Y , the pre-image f −1 (V ) is open in X.
Proof. Suppose f is continuous, and V ⊂ Y is open. Then ∀x ∈ f −1 (V ), by Proposition
1.4, f −1 (V ) contains an open set U with x ∈ U. So f −1 (V ) is open in X.
Conversely suppose for any open set V ⊂ Y , the pre-image f −1 (V ) is open in X.
For any x ∈ X, take any open set V in Y with f (x) ∈ V . Then f −1 (V ) itself is an
open set in X which contains the point x. So by Proposition 1.4, f is continuous.
Definition 1.7 (Topologically equivalent metrics). Let d1 and d2 be two metrics on a
set X. We say d1 and d2 are topologically equivalent if they produce the same set of
open sets.
Obviously strongly equivalent metrics are always topologically equivalent, but the
converse is not true. In general, we will call a conception a “topological conception”
if the conception depends only on the collection of open sets (this will be clear later).
So “neighborhood” is a topological conception, i.e. it depends only on the collection
of open sets, and continuity is a topological property.
As a consequence of Theorem 1.6, we have
Corollary 1.8. Suppose dÝ Ý
X and dY are metrics that are topologically equivalent to
dX and dY respectively, then f : (X, dX ) → (Y, dY ) is continuous if and only if f :
(X, dÝ Ý
X ) → (Y, dY ) is continuous.
This is why d1 , d2 , d¯2 on Rn produce the same set of continuous functions, while the
discrete metric produce a different set: it is not hard to see from the arguments in the
examples above that the collection of open sets determined by d1 , d2 or d¯2 are all the
same, while the collection of open sets determined by the discrete metric is different!
In conclusion:
Although we defined continuity via the metric structure, continuity is
really a conception that depends only on the collection of open sets
produced by the metric!
¶ A non-topological conception: the uniform continuity.
To compare, let’s take a look at a similar conception: the “uniform continuity” of
a map between metric spaces. The definition is straightforward:
4 TOPOLOGY: DEFINITIONS AND EXAMPLES
Definition 1.9. A map f : (X, dX ) → (Y, dY ) is uniformly continuous if
∀ε > 0, ∃δ > 0, s.t. dX (x1 , x2 ) < δ ⇒ dY (f (x1 ), f (x2 )) < ε.
Of course uniformly continuous functions are continuous, but the converse is not
true. It turns out that “uniform continuity” is NOT a topological property: it does
depend on the metric.2
Example 1.10. Let d be the standard metric on R, and let d1 be the metric on R
induced by the map arctan : R → (− π2 , π2 ), i.e.
d1 (x, y) := | arctan(x) − arctan(y)|.
Then open balls of d1 are exactly open intervals in R. So d and d1 induces the same
set of open sets, i.e. they are topologically equivalent.
Consider the identity map
f : R → R, f (x) = x.
Then f : (R, d) → (R, d) is uniformly continuous, but f : (R, d1 ) → (R, d) is NOT
uniformly continuous since
d1 (n, n + 1) = | arctan(n) − arctan(n + 1)| → 0 as n → ∞
but d(n, n + 1) = 1.
2. Topology: Definitions and Examples
¶ Topology via neighborhood structure.
To extend the conceptions of continuity and convergence to more general “spaces”,
intuitively one need to axiomatize the conception of “neighborhood” first. Here is how
to do this:
For any x ∈ X, one can assign to it a non-empty collection of subsets,
x 7→ N (x) ⊂ P(X), 3
with the understanding that each element in N (x) is a neighborhood of x. The axioms
for these N (x)’s to satisfy are the following:
(N1) If N ∈ N (x), then x ∈ N.
(N2) If M ⊃ N and N ∈ N (x) then M ∈ N (x).
(N3) If N1 , N2 ∈ N (x), then N1 ∩ N2 ∈ N (x).
(N4) If N ∈ N (x), then ∃M ⊂ N and M ∈ N (x), s.t. ∀y ∈ M, N ∈ N (y).
2
There is a generalization of metric structure, called the “uniform structure”. One can define
uniform continuity for maps between spaces with uniform structures. For details, c.f. J.L. Kelley,
General Topology.
3We use the notation P(X), or sometimes 2X , to denote the power set of X, i.e. the set of all
subsets of X.
TOPOLOGY: DEFINITIONS AND EXAMPLES 5
Remarks 2.1.
(1) The first three axioms on neighborhood have clear meanings. The fourth one,
(N4), gives a relation between neighborhoods of different points and can be
regarded as a replacement of triangle inequality for the metric structure.
(2) Such a structure was first introduced by Hausdorff in 1912. 4 His goal was
to define a very general notion of space that includes Rn , Riemann surfaces,
infinitely dimensional spaces, or spaces consisting of curves or functions. He
gave two advantages of introducing such a general notion: to simplify theories,
and to prevent us from “illegally” using intuition.
Definition 2.2. A neighborhood structure N on a set X is a map
N : X → P(P(X)) \ {∅}
satisfying axioms (N1)-(N4).
One can call an abstract set X together with a neighborhood structure N a (neigh-
borhood) topological space. Although it looks very complicated, it turns out that it is
equivalent to the usual definition of topological space via open sets that we will give
below.
¶ Topology via interior structure.
Given a neighborhood structure (X, N ), how do we get the conception of open
sets in X? Recall that in mathematical analysis, a set is open if and only if every point
in the set is an interior point of the set. We can first define the conception of “interior”
in neighborhood topological space:
Definition 2.3. Let N be a neighborhood structure on X. For any subset A ⊂ X,
its interior is defined to be
Int(A) := {x ∈ A | A ∈ N (x)}.
Using definition and axioms (N1)-(N4) one can easily check
(I 1) Int(A) ⊂ A.
(I 2) Int(A) ∩ Int(B) = Int(A ∩ B).
(I 3) Int(Int(A)) = Int(A).
(I 4) Int(X) = X.
It turns out that the “interior structure”, i.e. a map
I : P(X) → P(X)
4
However, the axioms that Hausdorff proposed are a little bit different from the ones above: He
requires an additional separation axiom that if x 6= y, then there exists N ∈ N (x) and M ∈ N (y)
so that N ∩ M = ∅. Such a separation axiom is called Hausdorff property and will be studied later in
this course.
6 TOPOLOGY: DEFINITIONS AND EXAMPLES
satisfying (I 1)-(I 4) is also a structure which is equivalent to the neighborhood struc-
ture.
Both “neighborhood axioms” (N1)-(N4) and “interior axioms” (I 1)-(I 4) can be
used to define topology on a set. Although it is conceptionally easier to understand
the relation between neighborhoods and continuity, neighborhood axioms or interior
axioms are not easy to use in practice. The much-easier-to-use way to define topology
is via open sets.
¶ Topology via open sets.
Motivated by the conception of open sets in metric spaces (as sets all of whose
points are interior points), and given the conception of “interior” above, it is not hard
to give the following definition of open sets in (X, N ):
Definition 2.4. In a neighborhood topological space (X, N ), a set U is said to be
open if U ∈ N (x) for any x ∈ U , or equivalently, if Int(U ) = U .
Given (X, N ), if we let
T = {U ⊂ X | U is open}
be the set of all open sets in (X, N ), one can check:
(O1) ∅ ∈ T , X ∈ T .
(O2) If U1 , U2 ∈ T , so is U1 ∩ U2 .
(O3) If {Uα : α ∈ Λ} ⊂ T , then ∪α∈Λ Uα ∈ T .
Conversely, given a collection T satisfying (O1)-(O3), if for any x ∈ X we let
N (x) = {N ⊂ X : ∃U ∈ T s.t. x ∈ U and U ⊂ N },
then one can check: N (x) satisfies (N1), (N2), (N3), (N4). In today’s PSets you will
be asked to prove the equivalence between (O1)-(O3) and (N1)-(N4) [which is more
subtle than proving “(O1)-(O3) ⇒ (N1)-(N4)” and “(O1)-(O3) ⇒ (N1)-(N4)”].
Since the axioms (O1)-(O3) are much easier to use, we take them as the definition
of a topology:5
Definition 2.5 (Topology). A topology on a set X is a collection T ⊂ P(X) of subsets
of X which satisfies (O1), (O2) and (O3). We call such a pair (X, T ) a topological
space.
Since the conception of neighborhoods is so important, we defined it in a topological
space formally as follows:
Definition 2.6. Let (X, T ) be a topological space. A set N ⊂ X is called a neigh-
borhood of x if there exists an open set U ∈ T so that x ∈ U ⊂ N .
5Thisdefinition of topology via open sets first appeared in the book “Topology” written by Alexan-
droff and Hopf in 1935.
TOPOLOGY: DEFINITIONS AND EXAMPLES 7
¶ Topology via closed sets.
Of course, given the conception of open sets, one can define closed sets:
Definition 2.7. A set F in a topological space (X, T ) is closed if its complementary
F c = X \ F is open.
It is a trivial exercise to convert the “open sets axioms” to “closed sets axioms”:
(C1) Both ∅ and X are closed.
(C2) If U1 , U2 are closed, so is U1 ∪ U2 .
(C3) If Uα are closed for all α ∈ Λ, so is ∩α∈Λ Uα .
¶ Examples of topological spaces.
In what follows we give some examples of topologies.
Example 2.8. (The metric topology) Let (X, d) be any metric space. Let
Tmetric = {U ⊂ X | ∀x ∈ U, ∃r > 0 s.t. B(x, r) ⊂ U }.
Then Tmetric is a topology on X. It is called the metric topology.
Example 2.9. (The trivial and discrete topologies) Let X be any set. On X one can
always define two “extremal” topologies:
(1) The discrete topology
Tdiscrete = P(X) = {Y | Y ⊂ X}.
It is the metric topology associated to the discrete metric on X.
(2) The trivial topology (also called the “indiscrete topology”)
Ttrivial = {∅, X}.
It is NOT a metric topology for any X with more than one element.
Note for any topology T on X, we always have
Ttrivial ⊂ T ⊂ Tdiscrete
Definition 2.10. Let T1 and T2 be two topologies on X. we say T1 is weaker than 6
T2 , or equivalently, T2 is stronger than T1 , if T1 ⊂ T2 .
It follows that on any set X, Ttrivial is the weakest topology, while Tdiscrete is the
strongest topology.
Remark 2.11. It is possible that two different topologies on X are NOT comparable.
For example, the Euclidean topology and the cocountable topology in next example on
R are not comparable.
6Some authors use the word coarser than instead of “weaker than”, and use the word finer than
instead of “stronger than”.
8 TOPOLOGY: DEFINITIONS AND EXAMPLES
Example 2.12. (The cofinite and cocountable topologies) Let X be any infinite set.
(1) The cofinite topology
Tcof inite = {A ⊂ X | A = ∅ or Ac = X \ A is a finite set}.
• ∅ ∈ T ; X ∈ T since X c = ∅ is finite.
• If A, B ∈ T , and A, B 6= ∅. Then Ac , B c are finite. So (A∩B)c =
Ac ∪ B c is finite.
• If Aα ∈ T and at least one Aα1 6= ∅, then (∪α Aα )c = ∩α Acα ⊂
Acα1 is finite.
(2) Similarly one can define the cocountable topology (for any set X with uncount-
ably many elements):
Tcocountable = {A ⊂ X | A = ∅ or Ac is at most countable }.
[Please check that it is a topology]
Example 2.13. (The Zariski topology) Let X = Cn . Let R = C[z1 , · · · , zn ], i.e. the
polynomial ring of n-variables with complex coefficients. Define
TZariski = {U ⊂ Cn | ∃f1 , · · · , fm ∈ R s.t. U c = common zeroes of f1 , · · · , fm }.
One can prove that it is a topology. More generally one can define the Zariski topology
on any ring. algebraic geometry.
Example 2.14. (The Sorgenfrey topology) Let X = R. Consider
TSorgenf rey = {U ⊂ R | ∀x ∈ U, ∃ε > 0 s.t. [x, x + ε) ⊂ U }.
Then one can check that it is a topology.
¶ New topological spaces from old.
As in the case of abstract metric spaces, one can construct new topologies from old
ones via the standard “restriction to subset” operation and via the “Cartesian product”
operations:
(1) (The subspace topology) Let (X, T ) be a topological space, and Y ⊂ X a
subset. Then the collection
TY := {U ∩ Y | U ∈ T }
form a topology on Y . It is called the subspace topology.
Remark 2.15. If (X, dX ) is a metric space and Y ⊂ X, then “the subspace
topology on Y induced from the metric topology on X” coincides with “the
metric topology of (Y, dY ), where dY is the induced subspace metric”.
(2) (The product topology) Let (X, TX ) and (Y, TY ) be topological spaces. Then
TX×Y := {W ⊂ X×Y | ∀(x, y) ∈ W, ∃U ∈ TX and V ∈ TY s.t. (x, y) ∈ U ×V ⊂ W }
is a topology on X × Y. It is called the product topology.
TOPOLOGY: DEFINITIONS AND EXAMPLES 9
Remark 2.16. For metrics spaces, “the metric topologies induced from the var-
ious product metrics on X × Y (c.f. Lecture 2)” are all the same, and coincide
with “the product topology of the metric topologies on each component”!