8.5 Topological spaces

A topological space is the most general setting where you can talk about continuity, connectedness, and convergence without needing a notion of distance. By stripping away the specifics of distance and coordinates, topology lets you focus on the structural relationships between points and sets, revealing deep connections across many areas of mathematics.

This section covers the axioms that define topological spaces, the key properties used to classify them, and the constructions and tools that make topology such a powerful framework.

Fundamentals of topological spaces

Topological spaces generalize familiar settings like the real number line by replacing the concept of distance with a more abstract structure built from open sets. The result is a framework flexible enough to study continuity and convergence in contexts where metrics don't exist or aren't useful.

Definition and basic concepts

A topological space is a pair $(X, \mathcal{T})$ where $X$ is a set and $\mathcal{T}$ is a collection of subsets of $X$ (called open sets) satisfying three axioms:

1. The empty set $\emptyset$ and the whole set $X$ are both in $\mathcal{T}$.
2. The intersection of any finite number of open sets is open (i.e., in $\mathcal{T}$).
3. The union of any arbitrary collection of open sets (finite, countable, or uncountable) is open.

The collection $\mathcal{T}$ is called a topology on $X$. It determines which subsets count as "open" and thereby defines the structure of nearness and continuity on the space.

A homeomorphism is a bijective continuous function whose inverse is also continuous. Two spaces related by a homeomorphism are considered topologically identical, meaning any property preserved by continuous deformation (a topological invariant) is the same for both.

Open and closed sets

• Open sets are the elements of $\mathcal{T}$. They form the building blocks of the topology and define the "nearness" structure of the space.
• A closed set is the complement of an open set. Equivalently, a set is closed if and only if it contains all of its limit points.
• The interior of a set $A$ is the largest open set contained in $A$: the collection of all points that have some open neighborhood entirely within $A$.
• The exterior of $A$ consists of points that have an open neighborhood entirely outside $A$ (disjoint from $A$).
• A boundary point of $A$ is a point where every open neighborhood intersects both $A$ and its complement. Boundary points belong to the closure of $A$ but not its interior.

Neighborhoods and interior points

A neighborhood of a point $x$ is any set that contains an open set which itself contains $x$. Neighborhoods capture the local structure around a point.

• Interior points of a set $A$ have at least one neighborhood entirely contained in $A$. The collection of all interior points forms the largest open subset of $A$, written $\text{int}(A)$.
• Exterior points of $A$ have a neighborhood entirely outside $A$, placing them in $\text{int}(X \setminus A)$.
• An accumulation point (or limit point) of $A$ is a point where every neighborhood contains at least one point of $A$ other than itself. An accumulation point may or may not belong to $A$.

Boundary and closure

• The boundary $\partial A$ consists of all points that are neither interior to $A$ nor interior to its complement. Every neighborhood of a boundary point "straddles" the edge of $A$.
• The closure $\overline{A}$ is the smallest closed set containing $A$. It equals $A$ together with all its limit points: $\overline{A} = A \cup A'$, where $A'$ is the derived set.
• The derived set $A'$ is the set of all accumulation points of $A$. It can be a proper subset of the closure since $\overline{A}$ also includes isolated points of $A$.
• A set is dense in $X$ if its closure equals $X$. For example, the rationals $\mathbb{Q}$ are dense in $\mathbb{R}$ because every open interval of real numbers contains a rational.
• A set is nowhere dense if the interior of its closure is empty. The Cantor set in $\mathbb{R}$ is a classic example: despite being uncountable, it contains no interval.

Properties of topological spaces

Topological properties are those that remain unchanged under homeomorphism. They characterize a space independently of any particular way you might represent or embed it.

Connectedness

A space is connected if it cannot be written as the union of two disjoint, non-empty open sets. Intuitively, a connected space is "all one piece."

• Path-connected is a stronger condition: for any two points, there exists a continuous path (a continuous function from $[0,1]$) joining them. Path-connected always implies connected, but not vice versa.
• Components are the maximal connected subsets of a space. They partition the space into its "pieces."
• A space is simply connected if it is path-connected and every loop can be continuously shrunk to a point (no holes). The sphere $S^2$ is simply connected; the torus is not.
• A space is locally connected if every point has a neighborhood basis of connected sets.

Compactness

A space is compact if every open cover has a finite subcover. That is, whenever you cover the space with (possibly infinitely many) open sets, finitely many of them already do the job.

Compactness generalizes finiteness and ensures that many useful properties of finite sets carry over. For instance, a continuous real-valued function on a compact space always attains its maximum and minimum.

• Sequentially compact: every sequence has a convergent subsequence. In metric spaces, this is equivalent to compactness, but in general topological spaces it is not.
• Locally compact: every point has a neighborhood whose closure is compact. The real line $\mathbb{R}$ is locally compact but not compact.
• Tychonoff's theorem: the product of any collection of compact spaces is compact (in the product topology). This is one of the most important results in general topology and is equivalent to the axiom of choice.

Separability

A space is separable if it contains a countable dense subset. The real line is separable because $\mathbb{Q}$ is countable and dense in $\mathbb{R}$.

• A space is second-countable if its topology has a countable basis (a countable collection of open sets from which all open sets can be built via unions).
• In metric spaces, separability and second-countability are equivalent. In general topological spaces, second-countability is strictly stronger.
• A Lindelöf space has the property that every open cover admits a countable subcover. Second-countable spaces are always Lindelöf.
• Separability plays a key role in functional analysis and measure theory, where working with countable dense subsets simplifies many arguments.

Metrizability

A topological space is metrizable if there exists a metric (distance function) on the set that generates exactly the given topology.

• The Urysohn metrization theorem states that a second-countable, regular, Hausdorff space is metrizable.
• The Nagata–Smirnov metrization theorem gives a more general characterization: a space is metrizable if and only if it is regular, Hausdorff, and has a $\sigma$-locally finite basis.
• Metrizable spaces are always Hausdorff and paracompact.
• Non-metrizable spaces (like the long line or the Sorgenfrey plane) show that metric-based intuition can fail in general topology.

Types of topological spaces

Different types of topological spaces satisfy different "niceness" conditions. Understanding the hierarchy helps you know which theorems apply in which settings.

Hausdorff spaces

A Hausdorff space (also called a $T_2$ space) is one where any two distinct points can be separated by disjoint open neighborhoods. This is the separation condition most commonly assumed in analysis.

• The Hausdorff property guarantees that limits of convergent sequences (and nets) are unique.
• Regular Hausdorff spaces ($T_3$): you can separate a point from a closed set not containing it using disjoint open sets.
• Tychonoff spaces (completely regular Hausdorff): for any point and closed set not containing it, there exists a continuous function into $[0,1]$ separating them. These spaces embed into products of $[0,1]$.

Metric spaces vs topological spaces

Every metric space naturally becomes a topological space: the open sets are unions of open balls $B(x, \epsilon) = \{y \in X : d(x,y) < \epsilon\}$. But not every topological space arises this way.

• Metric spaces are always first-countable (countable local basis at each point) and Hausdorff.
• Topological spaces provide a strictly more general framework. Many spaces that arise naturally in algebra, logic, and functional analysis are not metrizable.
• The advantage of the topological framework is that it isolates exactly the structure needed for continuity and convergence, without the extra baggage of a distance function.

Discrete vs indiscrete topologies

These two extremes illustrate the range of possible topologies on a set:

• The discrete topology on $X$ declares every subset to be open. It is the finest (largest) topology. Every function out of a discrete space is continuous. Discrete spaces are metrizable using the discrete metric: $d(x,y) = 1$ if $x \neq y$, and $d(x,x) = 0$.
• The indiscrete (trivial) topology on $X$ has only $\emptyset$ and $X$ as open sets. It is the coarsest (smallest) topology. Every function into an indiscrete space is continuous. An indiscrete space with more than one point is not Hausdorff, since you can't separate any two distinct points.

Comparing these two extremes highlights how the choice of topology controls the degree of "separation" and the behavior of continuous functions.

Continuous functions

Continuous functions are the central objects of study in topology. They are the maps that respect topological structure, and understanding their behavior is the key to most topological reasoning.

Definition in topological context

A function $f: X \to Y$ between topological spaces is continuous if the preimage of every open set in $Y$ is open in $X$. Notice this is defined in terms of preimages (inverse images), not images.

Equivalent formulations:

• The preimage of every closed set in $Y$ is closed in $X$.
• For every point $x \in X$ and every neighborhood $V$ of $f(x)$, there exists a neighborhood $U$ of $x$ such that $f(U) \subseteq V$.

Two stronger forms of continuity appear frequently:

• Uniform continuity (in metric spaces): the "$\delta$" in the $\epsilon$-$\delta$ definition can be chosen independently of the point. This ensures continuity behaves evenly across the entire domain.
• Lipschitz continuity: there exists a constant $K$ such that $d(f(x), f(y)) \leq K \cdot d(x,y)$ for all $x, y$. This bounds how fast the function can change.

Homeomorphisms

A homeomorphism is a bijection $f: X \to Y$ where both $f$ and $f^{-1}$ are continuous. If a homeomorphism exists between two spaces, they are topologically equivalent: they share all topological properties.

• Homeomorphisms preserve connectedness, compactness, separation axioms, and every other topological invariant.
• Intuitively, a homeomorphism corresponds to stretching, bending, or twisting a space without tearing or gluing.
• Classic example: a coffee mug and a donut (torus) are homeomorphic, since each has exactly one "hole."

Open and closed maps

• An open map sends open sets to open sets (in the codomain).
• A closed map sends closed sets to closed sets.
• A continuous function is not necessarily open or closed. These are independent properties.
• A quotient map is a surjection where a set in the codomain is open if and only if its preimage is open. Quotient maps are central to constructing new spaces from old ones.
• A perfect map is a closed, continuous, surjective map with compact fibers (preimages of points).

Constructions in topology

New topological spaces can be built from existing ones using standard constructions. Each construction comes with rules about which properties are preserved and which may be lost.

Subspace topology

Given a topological space $(X, \mathcal{T})$ and a subset $A \subseteq X$, the subspace topology on $A$ consists of all sets of the form $A \cap U$ where $U \in \mathcal{T}$.

• Properties like Hausdorffness and metrizability are inherited by subspaces.
• Compactness and connectedness are not automatically inherited. For example, $(0,1) \cup (2,3)$ is a subspace of $\mathbb{R}$ that is neither connected nor compact, even though $\mathbb{R}$ is connected.
• Important examples: intervals in $\mathbb{R}$, the sphere $S^n$ as a subspace of $\mathbb{R}^{n+1}$.

Product topology

Given spaces $X$ and $Y$, the product topology on $X \times Y$ has a basis consisting of products $U \times V$ where $U$ is open in $X$ and $V$ is open in $Y$.

• The projection maps $\pi_X: X \times Y \to X$ and $\pi_Y: X \times Y \to Y$ are continuous and open.
• Hausdorffness is preserved. Compactness is preserved by Tychonoff's theorem.
• For infinite products, the product topology (only finitely many coordinates constrained at a time) differs from the box topology (all coordinates can be constrained independently). The product topology is almost always the "right" choice because it makes Tychonoff's theorem work.
• The Sorgenfrey plane (product of the Sorgenfrey line with itself) is a useful source of counterexamples: it is separable but not second-countable, and not metrizable.

Quotient topology

The quotient topology arises from an equivalence relation on a space. If $\sim$ is an equivalence relation on $X$, the quotient space $X/{\sim}$ has the topology where a set is open if and only if its preimage under the quotient map $q: X \to X/{\sim}$ is open in $X$.

• This construction is how you "glue" parts of a space together. For example, identifying opposite edges of a square produces a torus.
• Hausdorffness can fail in quotient spaces, so you often need to check this separately.
• The quotient map satisfies a universal property: a function out of the quotient space is continuous if and only if its composition with the quotient map is continuous.

Convergence in topological spaces

In metric spaces, convergence is defined using sequences and distance. In general topological spaces, sequences alone may not capture the full convergence behavior, so more general tools are needed.

Nets and filters

• A net is a generalization of a sequence where the index set is a directed set (not necessarily $\mathbb{N}$). Nets can detect convergence behavior that sequences miss in non-first-countable spaces.
• A filter is a collection of subsets satisfying certain axioms (closed under finite intersection and supersets, does not contain $\emptyset$). Filters provide an alternative, equivalent framework for describing convergence.
• In compact spaces, every universal net (ultranet) has a cluster point, paralleling the Bolzano–Weierstrass theorem for sequences in $\mathbb{R}$.

Sequential spaces

A sequential space is one whose topology is completely determined by its convergent sequences: a set is closed if and only if it is closed under sequential limits.

• Every first-countable space is sequential, but not every sequential space is first-countable.
• In sequential spaces, a function is continuous if and only if it preserves convergent sequences (sequential continuity equals continuity).
• Not all spaces are sequential. The Stone–Čech compactification of the integers, $\beta\mathbb{N}$, is a standard example of a non-sequential space.

First-countable spaces

A space is first-countable if every point has a countable neighborhood basis (a countable collection of neighborhoods such that every neighborhood of the point contains one from the collection).

• In first-countable spaces, sequences suffice to describe the topology: closures, continuity, and compactness can all be characterized using sequences.
• All metric spaces are first-countable.
• First-countable does not imply second-countable. For example, an uncountable discrete space is first-countable but not second-countable.

Separation axioms

The separation axioms form a hierarchy that classifies spaces by how well you can "pull apart" points and closed sets using open sets. Higher separation axioms give you more powerful tools but apply to fewer spaces.

$T_0$, $T_1$, and $T_2$ spaces

• $T_0$ (Kolmogorov): For any two distinct points, at least one has an open neighborhood not containing the other. This is the weakest separation axiom.
• $T_1$ (Fréchet): For any two distinct points, each has an open neighborhood not containing the other. Equivalently, every singleton $\{x\}$ is a closed set.
• $T_2$ (Hausdorff): Any two distinct points have disjoint open neighborhoods.

The hierarchy is strict: $T_0 \subsetneq T_1 \subsetneq T_2$. Most spaces encountered in analysis are at least Hausdorff. The $T_0$ axiom is important in domain theory and certain areas of computer science.

Regular and normal spaces

• Regular ($T_3$): A $T_1$ space where any point and any closed set not containing it can be separated by disjoint open neighborhoods.
• Normal ($T_4$): A $T_1$ space where any two disjoint closed sets can be separated by disjoint open neighborhoods.
• Urysohn's lemma: In a normal space, for any two disjoint closed sets $A$ and $B$, there exists a continuous function $f: X \to [0,1]$ with $f(A) = \{0\}$ and $f(B) = \{1\}$. This is a powerful bridge between topology and analysis.
• Tietze extension theorem: Any continuous real-valued function defined on a closed subset of a normal space can be extended to a continuous function on the whole space.
• Every paracompact Hausdorff space is normal. Since all metric spaces are paracompact, all metric spaces are normal.

Applications of topology

Topology provides tools and perspectives that reach far beyond the study of topological spaces themselves.

In analysis and geometry

• Functional analysis relies on topological vector spaces (Banach spaces, Hilbert spaces) where the topology comes from a norm or inner product.
• Differential topology studies smooth manifolds and smooth maps, using topological ideas to understand differentiable structures.
• Geometric topology focuses on low-dimensional manifolds (surfaces, 3-manifolds) and knot theory.
• Topological methods are essential in the study of partial differential equations, where solution spaces often carry natural topologies.

In algebraic topology

Algebraic topology assigns algebraic objects (groups, rings) to topological spaces, turning geometric problems into algebraic ones.

• Homotopy theory studies when two maps or spaces can be continuously deformed into each other.
• Homology and cohomology provide computable invariants that detect "holes" of various dimensions.
• The fundamental group $\pi_1(X)$ captures the loop structure of a space. For example, $\pi_1(S^1) \cong \mathbb{Z}$, reflecting the fact that loops around a circle are classified by how many times they wind around.
• Covering spaces are closely tied to the fundamental group and provide geometric insight into its algebraic structure.

In data science and computing

• Topological data analysis (TDA) uses tools from algebraic topology to extract shape information from high-dimensional datasets.
• Persistent homology tracks how topological features (connected components, loops, voids) appear and disappear across multiple scales, providing a robust summary of data shape.
• Network topology studies the connectivity and structural properties of complex networks.
• Topological ideas also appear in quantum computing and condensed matter physics, where topological invariants classify phases of matter.