Back to notes
Topology

mathtopology

Topology is the study of spaces and their properties.

Topological space

Let XX be a set. A topology over XX is some subset τP(X)\tau\subseteq \mathcal{P}(X) such that:

  1. ,Xτ\emptyset, X\in \tau.

  2. {Ui}iIτiIUiτ\{U_i\}_{i\in I}\subseteq \tau \Longrightarrow \bigcup_{i\in I} U_i\in\tau.

  3. {Ui}i=1,,nτi=1nUiτ\{U_i\}_{i=1,\ldots,n}\subseteq \tau \Longrightarrow \bigcap_{i=1}^{n} U_i\in\tau.

The subsets AτA\in\tau are called the open subsets of the topology.

A topological space is a set XX endowed with some specific topology τ\tau. Despite a topological space is technically a pair (X,τ)(X, \tau), we might as well refer to the "topological space XX" whenever the topology is understood, e.g. the topological space R\R (whose open sets are the clasical open sets of real analysis).

Initial topologies and embeddings

Definition.(Initial topologies)

Let f ⁣:XYf\colon X\to Y be a map. Suppose YY is endowed with some topology. We define the initial topology on XX with respect to ff as

τ={f1(U)Uτ}\tau = \{f^{-1}(U)\mid U\in \tau\}

This is, the coarsest topology on XX such that the map f ⁣:XYf\colon X\to Y is continuous.

We can suppose that XX and YY are already topological spaces and ask ourselves which conditions on ff guarantee that the map f ⁣:XYf\colon X\to Y is "initial", i.e., XX has the initial topology. By definition, if there're some x1,x2Xx_1, x_2\in X s.t. f(x1)=f(x2)f(x_1) = f(x_2) then the open neighborhoods of x1x_1 and x2x_2 are exactly the same. Thus, x1x_1 and x2x_2 are topologically indistinguishable (see Kolmogorov axiom). Thus, we can "collapse" the topologically equivalent points in the domain and get an initial and injective function ff. Being initial thus becomes equivalent to the fact that ff is a homeomorphism of XX and f(X)f(X), which is usually called being an embedding.

Definition.(Embeddings)

Let f ⁣:XYf\colon X\to Y be a continuous map between topological spaces. We say that ff is an embedding, or, equivalently, that XX is an (embedded) subspace of YY, if F:XYF: X\to Y is an homeomorphism between XX and f(X)f(X). Equivalently, a map ff is an embedding if and only if ff is injective and XX has the initial topology with respect to ff.

For example, let us consider the map f:[0,1)R2f: [0, 1)\to \R^2 given by f(t)=(cos2πt,sin2πt)f(t) = (\cos{2\pi t}, \sin{2\pi t}). If we take the usual topology if [0,1][0, 1], then this map is not an embedding since the open set [0,ε)[0,1)[0, \varepsilon)\subseteq [0, 1) cannot be the preimage of any open set in R2\R^2. This occurs as [0,1)[0, 1) is definitely not homeomorphic to the subspace S1R2S^1\subseteq \R^2.

If we take the mapping f(t)=(cosπt,sinπt)f(t) = (\cos{\pi t}, \sin{\pi t}) instead, this function is an embedding, as a semicircumference is homeomorphic to a segment.

Otherwise, we could instead choose the initial topology on [0,1)[0, 1) given by the former mapping. With this topology [0,ε)[0, \varepsilon) is not an open set, as in fact every neighborhood of 00 must contain 1ε1-\varepsilon for some ε\varepsilon. This is like giving the "S1S^1 topology" to the semiopen interval, which entails "gluing" the endpoints together.

Hausdorff axiom

Our current definition of topological space has some drawbacks. There're some "intuitive" facts that do not hold anymore for arbitrary topological spaces:

  1. A singleton {x}\{x\} does not need to be a closed subset.

  2. A sequence (xn)n1(x_n)_{n\ge 1} can have more than one limit.

(Prove that if every sequence has a unique limit then the set {x}\{x\} must be closed.)

There is an additional axiom which allows us to prove the previous two facts for every topological space that satisfies it.

Definition.(Hausdorff axiom)

Let XX be a topological space. We say that XX satisfies the Hausdorff axiom (also called the T2T_2 axiom) if for every x,yXx, y\in X there're open sets UU, VV such that xUx\in U, yVy\in V and UV=U\cap V = \emptyset.

Kolmogorov axiom

We now study an axiom which is strictly weaker than the Hausdorff axiom.

Definition.(Kolmogorov)

Let XX be a topological space. We say that XX satisfies the Kolmogorov axiom (also called the T0T_0 axiom) if for every x,yXx, y\in X there is an open set UU such that xUx\in U and y∉Uy\not\in U or x∉Ux\not\in U and yUy\in U.

It is remarkable that all spaces satisfying this axiom can be characterized in a very specific way.

Compactness

Definition.(compact spaces)

Let XX be a topological space. We say that XX is compact if every open cover of XX admits a finite subcover.

Theorem.(equivalences of compactness)

The following are equivalent definitions of compactness over an arbitrary topological space XX:

  1. every open cover of XX admits a finite subcover.
  2. every family of open sets with the FIP has non-empty intersection.
  3. every net has a convergent subnet.
  4. the function f:X×ZXf: X\times Z\to X is closed for every space ZZ.

The equivalence of 1. and 2. is evident.

To prove that 1. implies 3., we consider an arbitrary net (xλ)λΛ(x_\lambda)_{\lambda\in \Lambda}. Suppose the net doesn't have a convergent subnet. Then, for every xXx\in X, there is some open neighbourhood UxU_x such that no subnet is eventually contained in UxU_x. This means that the set Ax={λΛxλUx}A_x = \{\lambda\in\Lambda\mid x_\lambda\in U_x\} is not cofinal, which means that there is some λxΛ\lambda_x\in \Lambda such that there is no αAx\alpha\in A_x with λxα\lambda_x\le \alpha. As every open cover has a finite subcover, there is a finite set x1,x2,,xnx_1, x_2, \ldots, x_n such that their corresponding open sets cover all XX. If we take λ0λx1,,λxn\lambda_0 \ge \lambda_{x_1}, \ldots, \lambda_{x_n}, then there cannot be any λλ0\lambda\ge \lambda_0 in Λ\Lambda. As Λ\Lambda is directed, this means that λ0\lambda_0 is an upper bound, and therefore xλ0x_{\lambda_0} must be an accumulation point of the net, a contradiction.

To prove that 3. implies 2., we consider an arbitrary family (Fi)iI(F_i)_{i\in\mathcal{I}} with the FIP. Thus, for every finite JI\mathcal{J}\subseteq\mathcal{I} we have some xJjJFjx_\mathcal{J}\in \bigcap_{j\in\mathcal{J}}{F_j}. This gives a net (xJ)(x_{\mathcal{J}}) directed by the poset of finite subsets of I\mathcal{I}, which must have a convergent subnet. Let x0x_0 be the limit point, x0Ux_0\in U an arbitrary neighbourhood and iIi\in \mathcal{I} an arbitrary index. Because every subnet is cofinal, there is some J{i}\mathcal{J}\supseteq \{i\} such that xJUx_\mathcal{J}\in U, which means that FiUF_i\cap U\neq \varnothing. As this assertion is true for every UU, then x0x_0 is a limit point of FiF_i, which is closed, and thus x0Fix_0\in F_i. As this is true for every iIi\in\mathcal{I}, we have that x0iIFix_0\in\bigcap_{i\in\mathcal{I}}{F_i}, as desired.

Definition.(proper maps)

A continuous map f ⁣:XYf\colon X\to Y is proper if it is closed and its fibers are compact.

Theorem.

The following definitions of proper map are equivalent:

  1. it is closed and its fibers are compact.
  2. preimage of compact sets is compact.
Theorem.

Composition of proper maps is proper.

We can actually put the concept of compactness into a even more general setting, which involves using pullbacks and universally closed maps.

Definition.(fibered products)

Let XX, YY, ZZ be topological spaces and f:XZf: X\to Z, g:YZg: Y\to Z continuous maps. Then, the fibered product of XX and YY through ff and gg is the space

X×Y={(x,y)X×Yf(x)=g(y)}X\times Y = \left\{(x, y)\in X\times Y\mid f(x) = g(y)\right\}

with the subspace topology.

Definition.(pullbacks)

Let XX, YY, ZZ be topological spaces and f:XZf: X\to Z, g:YZg: Y\to Z continuous maps. Then, the pullback of XX and YY through ff and gg is the unique space WW, endowed with projections f ⁣:WY\overline{f}\colon W\to Y and g ⁣:WX\overline{g}\colon W\to X such that fg=gff\overline{g} = g\overline{f} with the following universal property: for every other space WW' endowed with projections f ⁣:WYf'\colon W'\to Y and g ⁣:WXg'\colon W'\to X such that fg=gffg' = gf', there is a unique continuous function φ ⁣:WW\varphi\colon W'\to W such that f=φff' = \varphi \overline{f} and g=φgg' = \varphi \overline{g}.

This is, given an "incomplete" commutative square, the pullback is the missing corner which not only makes the diagram commute, but it is also the closest object which makes the diagram commute.

Definition.(universally closed maps)

A continuous map f ⁣:XYf\colon X\to Y is universally closed if for every topological space ZZ, the continuous map f×idZ ⁣:X×ZY×Zf\times\text{id}_Z\colon X\times Z\to Y\times Z is closed.

Theorem.

The following definitions of universally closed map are equivalent:

  1. for every topological space ZZ, the continuous map f×idZ ⁣:X×ZY×Zf\times\text{id}_Z\colon X\times Z\to Y\times Z is closed.
  2. for every topological space ZZ, if WW is the pullback of XX and ZZ through YY with projection f ⁣:WZ\overline{f}\colon W\to Z, then f\overline{f} is closed.

The implication from 2. to 1. is trivial.

Theorem.

Given a continuous function f ⁣:XYf\colon X\to Y, then ff is proper if and only if it is universally closed.