Topological space

A topological space is a pair $(X, \tau)$ consisting of a set $X$ (the underlying space) and a collection $\tau$ of subsets of $X$ (called open sets), where $\tau$ satisfies the following axioms: the empty set $\emptyset$ and $X$ itself belong to $\tau$; $\tau$ is closed under arbitrary unions; and $\tau$ is closed under finite intersections.^[1] This structure axiomatizes the notion of continuity and nearness in a general way, without relying on distances or coordinates, making it more abstract than metric spaces but applicable to a wide range of geometric and analytic contexts.^[2] The concept originated in the early 20th century as part of efforts to formalize notions of continuity and limit points in set theory. Felix Hausdorff introduced an early version in his 1914 book Grundzüge der Mengenlehre, where he defined a topological space using neighborhoods with a separation axiom (now known as the Hausdorff condition) built in as a requirement.^[1] Kazimierz Kuratowski generalized this in 1922 by providing closure axioms that define a topology without mandating separation, allowing for broader classes of spaces including non-Hausdorff examples.^[1] The modern open-set definition, equivalent to Kuratowski's, was further popularized by the Bourbaki group in the 1940s, solidifying its role as the standard foundation for topology.^[1] Topological spaces form the core of point-set topology, enabling the study of properties like compactness, connectedness, and separation that are invariant under homeomorphisms (continuous bijections with continuous inverses).^[3] They unify diverse areas of mathematics: in analysis, they generalize Euclidean spaces for convergence and uniform structures; in geometry, they underpin manifolds and differential topology; in algebraic topology, they support homotopy and homology theories for classifying shapes up to deformation.^[4][5] Beyond pure math, topological ideas appear in physics (e.g., configuration spaces in quantum mechanics) and data science (e.g., persistent homology for shape analysis in high-dimensional data).^[6]

History

Early developments

The concept of topological spaces originated from intuitive geometric and analytic ideas in the 19th century, particularly through efforts to understand continuity and connectivity without reliance on metrics. In his 1851 doctoral dissertation, Bernhard Riemann introduced Riemann surfaces as a means to handle multi-valued inverse functions in complex analysis, such as the square root or logarithm, by constructing a branched covering space over the complex plane where these functions become single-valued and continuous. This approach emphasized the intrinsic geometry of the surface itself, independent of embedding in Euclidean space, and highlighted properties like connectivity and branching that anticipated abstract topological structures.^[7] Building on such geometric insights, mathematicians began exploring separation and domain properties in the plane. In 1887, Camille Jordan established the Jordan curve theorem, proving that every simple closed curve in the Euclidean plane divides it into exactly two connected regions: a bounded interior and an unbounded exterior, with no path connecting points from one region to the other without crossing the curve. This result provided a foundational understanding of how curves organize plane domains and influenced subsequent work on continuity in non-metric settings. Around the turn of the century, Arthur Schoenflies extended these ideas through his studies of plane topology, introducing notions like accessible points and simple closed curves, and proving the topological invariance of dimension for the square; his contributions clarified the structure of plane domains bounded by Jordan curves, showing that such interiors are homeomorphic to open disks.^[8][9] A pivotal advancement came in 1895 with Henri Poincaré's seminal paper "Analysis Situs," which shifted focus to combinatorial aspects of spaces. Poincaré developed methods to classify surfaces and higher-dimensional manifolds using invariants like genus—measuring the number of "handles" on a surface—and connectivity via what would later be recognized as homology groups, allowing qualitative distinctions between spaces based on their "holes" rather than measurements. This work formalized intuitive notions of topological equivalence for orientable and non-orientable surfaces, laying groundwork for algebraic tools to study spatial properties.^[10] These developments culminated in early 20th-century forums, such as the 1904 International Congress of Mathematicians in Heidelberg, where lectures and discussions on set-theoretic continua and point configurations—exemplified by talks on Cantor's continuum hypothesis—highlighted emerging interests in point-set approaches that would soon lead to axiomatic topology.^[11]

Formalization and key contributors

The formalization of topological spaces emerged in the early 20th century as mathematicians transitioned from concrete geometric and metric-based concepts to abstract set-theoretic frameworks, building on advances in set theory and influenced by Emmy Noether's emphasis on abstract algebraic structures and their invariances across mathematical disciplines. Noether's work, particularly her promotion of ideal theory and symmetry principles, encouraged a rigorous, axiomatic approach that extended to topology, fostering the development of general spaces independent of specific metrics or embeddings. Preceding Hausdorff, Maurice Fréchet introduced abstract notions of spaces in 1906, and L.E.J. Brouwer advanced point-set topology from 1911, laying groundwork for axiomatic definitions.^[12] A cornerstone of this formalization was Felix Hausdorff's 1914 monograph Grundzüge der Mengenlehre, which introduced a system of neighborhood axioms to define topological structures on sets, focusing on filters of neighborhoods to capture limits and continuity without relying on distances.^[13][14] Hausdorff's axioms, applied initially to what are now called Hausdorff spaces, included separation properties and laid the groundwork for abstract topology by treating spaces as collections of points with specified "nearness" relations derived from set theory.^[13] In the 1920s, Kazimierz Kuratowski advanced this axiomatization through his 1920 doctoral thesis, where he formulated four closure axioms that equivalently define topological spaces, allowing for a broader class of structures beyond Hausdorff's separation assumptions and paving the way for the modern open set formulation.^[15] These axioms used the closure operator to specify how sets accumulate points, providing a dual perspective to neighborhood systems and influencing subsequent refinements.^[15] Pavel Urysohn contributed crucially around 1922 (published 1925) with the Urysohn metrization theorem, which proved that every second-countable regular Hausdorff space is metrizable, thereby connecting abstract topological properties to uniform structures and metric spaces.^[16] This result, published in a series of notes, highlighted the interplay between topological axioms and uniformity, establishing key criteria for when general topologies admit compatible metrics.^[16] During the 1930s, Pavel Alexandrov and Heinz Hopf further refined these foundations in their 1935 textbook Topologie, which adopted and standardized the open set axioms as the primary definition of topological spaces, integrating prior neighborhood and closure approaches into a cohesive framework for general topology.^[17][18] Their work emphasized simplicial approximations and combinatorial methods, solidifying topology as an independent discipline while incorporating metrization results like Urysohn's to explore broader classes of spaces.^[17]

Definitions

Open sets

A topological space is formally defined as a pair $(X, \tau)$, where $X$ is a set and $\tau$ is a collection of subsets of $X$, called open sets, that satisfies certain axioms ensuring the structure captures intuitive notions of continuity and proximity. The collection $\tau$ is a subset of the power set $\mathcal{P}(X)$, and the elements of $\tau$ are the open sets of the topology. The axioms for $\tau$ to form a topology on $X$ are as follows:

1. The empty set $\emptyset$ and the whole set $X$ are in $\tau$.
2. The union of any arbitrary collection of sets in $\tau$ is in $\tau$.
3. The intersection of any finite collection of sets in $\tau$ is in $\tau$.
   These axioms ensure that $\tau$ is closed under arbitrary unions and finite intersections, providing a framework robust enough to model limits and convergence without relying on distances.

For any subset $A \subseteq X$, the interior of $A$, denoted $\operatorname{int}(A)$, is the largest open set contained in $A$, defined as the union of all open sets $U \in \tau$ such that $U \subseteq A$. This operation captures the "open core" of $A$ within the topology, and it is the unique largest such set because $\tau$ is closed under arbitrary unions. These axioms underpin the definition of continuous functions between topological spaces: a function $f: (X, \tau_X) \to (Y, \tau_Y)$ is continuous if the preimage $f^{-1}(V)$ is open in $X$ for every open set $V$ in $Y$. To see why the axioms suffice, note that the preimage operation preserves unions and intersections: $f^{-1}(\bigcup V_i) = \bigcup f^{-1}(V_i)$ for arbitrary families and $f^{-1}(\bigcap V_i) = \bigcap f^{-1}(V_i)$ for finite families, ensuring that if $\tau_Y$ satisfies the axioms, then so do the preimages under $f$, mirroring the structure on $X$. This preservation directly ties the open set axioms to the classical $\epsilon$-$\delta$ continuity in metric spaces, generalizing it to arbitrary topologies.

Neighborhoods

In a topological space $(X, \tau)$, a neighborhood of a point $x \in X$ is any subset $N \subseteq X$ that contains an open set $U \in \tau$ such that $x \in U$. Topological spaces can be defined equivalently using systems of neighborhoods rather than collections of open sets. For a set $X$, assign to each point $x \in X$ a family $\mathcal{N}(x)$ of subsets of $X$, called the neighborhoods of $x$. These families must satisfy the following axioms:^[19]

• (N1) $x \in N$ for every $N \in \mathcal{N}(x)$.
• (N2) If $N_1, N_2 \in \mathcal{N}(x)$, then there exists $N_3 \in \mathcal{N}(x)$ such that $N_3 \subseteq N_1 \cap N_2$.
• (N3) If $N \in \mathcal{N}(x)$ and $M \in \mathcal{N}(y)$ with $z \in N \cap M$, then there exists $N_z \in \mathcal{N}(z)$ such that $N_z \subseteq N \cap M$.

The pair $(X, \{\mathcal{N}(x) \mid x \in X\})$ satisfying these axioms defines a topological space.^[19] To recover the collection of open sets $\tau$ from the neighborhood systems, define $U \subseteq X$ as open if, for every $x \in U$, there exists $N \in \mathcal{N}(x)$ such that $N \subseteq U$. This construction yields a topology on $X$: the empty set and $X$ are open, arbitrary unions of open sets are open, and finite intersections of open sets are open. Conversely, given a topology $\tau$, the neighborhoods of $x$ can be taken as all subsets containing some open set from $\tau$ that contains $x$; this satisfies the neighborhood axioms. These two approaches are equivalent, as the open sets derived from neighborhoods match the original topology and vice versa.^[20][19] The neighborhood system $\mathcal{N}(x)$ at each point $x$ forms a filter on $X$, specifically the neighborhood filter, consisting of all sets containing some member of $\mathcal{N}(x)$. A filter base (or fundamental system of neighborhoods) for $\mathcal{N}(x)$ is a subfamily $\mathcal{B}(x) \subseteq \mathcal{N}(x)$ such that every neighborhood in $\mathcal{N}(x)$ contains some element of $\mathcal{B}(x)$, and $\mathcal{B}(x)$ is directed downward (any two elements have a third contained in their intersection). The concept of a filter base for the neighborhood system at a point is the local analogue of a basis for the entire topology (see ### Bases below). Filter bases provide a more economical way to specify the neighborhood system, as they generate the full filter while satisfying similar axioms, and are particularly useful in defining topologies with countable local bases (also called neighborhood bases at points), such as first-countable spaces.^[21][22]

Closed sets

A topological space $(X, \mathcal{C})$ can alternatively be defined by specifying a collection $\mathcal{C}$ of subsets of $X$, called closed sets, that satisfies the following axioms: the empty set ^[23] and the whole space $X$ belong to $\mathcal{C}$; the intersection of any (possibly infinite) subfamily of sets in $\mathcal{C}$ belongs to $\mathcal{C}$; and the union of any finite subfamily of sets in $\mathcal{C}$ belongs to $\mathcal{C}$.^[24] These axioms are the dual of those for open sets, obtained by interchanging unions and intersections and complementation. The closure operator associated with a topological space is a function $\mathrm{cl}: \mathcal{P}(X) \to \mathcal{P}(X)$, where $\mathcal{P}(X)$ is the power set of $X$, that assigns to each subset $A \subseteq X$ its closure $\mathrm{cl}(A)$, defined as the smallest closed set containing $A$ (i.e., the intersection of all closed sets containing $A$).^[24] This operator satisfies the Kuratowski closure axioms: $\mathrm{cl}(\emptyset) = \emptyset$; $A \subseteq \mathrm{cl}(A)$ (extensivity); $\mathrm{cl}(\mathrm{cl}(A)) = \mathrm{cl}(A)$ (idempotence); and $\mathrm{cl}(A \cup B) = \mathrm{cl}(A) \cup \mathrm{cl}(B)$ (preservation of finite unions).^[25] From these, monotonicity follows: if $A \subseteq B$, then $\mathrm{cl}(A) \subseteq \mathrm{cl}(B)$. These properties characterize the closure operator uniquely for any topology on $X$.^[24] Closed sets are precisely the complements of open sets: a subset $C \subseteq X$ is closed if and only if its complement $X \setminus C$ is open.^[26] Thus, the collection $\mathcal{C}$ is the set of all complements of sets in the open collection defining the topology. To see the equivalence of the open-set and closed-set axiomatizations, suppose $(X, \tau)$ is a topological space via open sets satisfying the standard axioms. Define $\mathcal{C} = \{X \setminus U \mid U \in \tau\}$. Then $\emptyset = X \setminus X \in \mathcal{C}$ and $X = X \setminus \emptyset \in \mathcal{C}$. For arbitrary intersections, if $\{C_\alpha\}_{\alpha \in I} \subseteq \mathcal{C}$, then $\bigcap_{\alpha \in I} C_\alpha = X \setminus \bigcup_{\alpha \in I} (X \setminus C_\alpha)$, and since $\bigcup (X \setminus C_\alpha) \in \tau$, the intersection is in $\mathcal{C}$. For finite unions, if $\{C_1, \dots, C_n\} \subseteq \mathcal{C}$, then $\bigcup_{i=1}^n C_i = X \setminus \bigcap_{i=1}^n (X \setminus C_i)$, and since the finite intersection of opens is open, the union is in $\mathcal{C}$. Conversely, starting from a closed collection $\mathcal{C}$ satisfying the dual axioms, the complements form an open collection satisfying the open axioms, establishing the bijection via complementation.^[24]

Bases

A basis (or base) for a topology on a set $X$ is a collection $\mathcal{B}$ of subsets of $X$ (called basis elements) satisfying the following conditions:

1. For every $x \in X$, there exists at least one $B \in \mathcal{B}$ such that $x \in B$.
2. If $x \in B_1 \cap B_2$ for $B_1, B_2 \in \mathcal{B}$, then there exists $B_3 \in \mathcal{B}$ such that $x \in B_3 \subseteq B_1 \cap B_2$.

A collection $\mathcal{B}$ satisfying these conditions generates a topology $\tau$ on $X$ consisting of all arbitrary unions of elements from $\mathcal{B}$, and $\mathcal{B}$ is then a basis for $\tau$. Equivalently, a basis $\mathcal{B}$ for a given topology $\tau$ is a subcollection $\mathcal{B} \subseteq \tau$ such that every open set in $\tau$ is a union of elements from $\mathcal{B}$.^[27][28] A subbasis (or subbase) for a topology is a collection $\mathcal{S}$ of subsets of $X$ such that the collection of all finite intersections of elements from $\mathcal{S}$ forms a basis for the topology. Any collection of subsets of $X$ can serve as a subbasis for some topology on $X$.^[28] Bases and subbases provide a convenient means to define topologies, particularly when the basis elements have a simple description and generate all open sets via unions (and finite intersections for subbases).

Equivalent formulations

A closure operator on a set $X$ provides an equivalent axiomatization of a topological space via the Kuratowski closure axioms. These axioms, introduced by Kazimierz Kuratowski, specify properties of a map $\mathrm{cl}: \mathcal{P}(X) \to \mathcal{P}(X)$, where $\mathcal{P}(X)$ denotes the power set of $X$:^[29]

$$\begin{align*} \mathrm{cl}(\emptyset) &= \emptyset, \\ A &\subseteq \mathrm{cl}(A) \quad \text{for all } A \subseteq X, \\ \mathrm{cl}(\mathrm{cl}(A)) &= \mathrm{cl}(A) \quad \text{for all } A \subseteq X, \\ \mathrm{cl}(A \cup B) &= \mathrm{cl}(A) \cup \mathrm{cl}(B) \quad \text{for all } A, B \subseteq X. \end{align*}$$

Such an operator defines a topology on $X$ by declaring a subset $C \subseteq X$ to be closed if $\mathrm{cl}(C) = C$, with open sets being the complements of closed sets.^[30] To prove equivalence to the open set definition, first note that if $(X, \mathcal{T})$ is a topological space, its closure operator $\mathrm{cl}(A) = \bigcap \{ F \subseteq X \mid A \subseteq F, F \text{ closed} \}$ satisfies the Kuratowski axioms: the empty set axiom follows from the fact that $X$ is closed; monotonicity from the intersection of closed sets containing $A$; idempotence from the closure of a closed set being itself; and additivity from the union of closed sets being closed. Conversely, given a Kuratowski closure operator, the family of fixed points $\{ C \subseteq X \mid \mathrm{cl}(C) = C \}$ consists of the closed sets, which form a topology dual to the open sets $\mathcal{T} = \{ X \setminus C \mid \mathrm{cl}(C) = C \}$: the empty set and $X$ are closed since $\mathrm{cl}(\emptyset) = \emptyset$ and $\mathrm{cl}(X) = X$ (by monotonicity and idempotence); arbitrary unions of closed sets are closed because if each $C_i$ satisfies $\mathrm{cl}(C_i) = C_i$, then $\mathrm{cl}(\bigcup C_i) = \bigcup \mathrm{cl}(C_i) = \bigcup C_i$; and finite intersections of closed sets are closed by De Morgan's laws applied to the dual open sets, using additivity and idempotence. The interior of a set $A \subseteq X$ is then given by $\mathrm{int}(A) = X \setminus \mathrm{cl}(X \setminus A)$, which recovers the open sets as interiors of themselves.^[31] Another equivalent formulation uses neighborhood systems. For each $x \in X$, a collection $\mathcal{N}_x \subseteq \mathcal{P}(X)$ of neighborhoods of $x$ satisfies:

1. $x \in N$ for every $N \in \mathcal{N}_x$,
2. if $N, M \in \mathcal{N}_x$, then there exists $N_3 \in \mathcal{N}_x$ such that $N_3 \subseteq N \cap M$,
3. if $N \in \mathcal{N}_x$, then there exists $M \in \mathcal{N}_x$ such that $M \subseteq N$,
4. if $N \in \mathcal{N}_x$, then for every $y \in N$, there exists $M_y \in \mathcal{N}_y$ with $M_y \subseteq N$.

These axioms ensure $\mathcal{N}_x$ forms a filter with the additional property that neighborhoods are "uniformly open" around their points. The corresponding topology has open sets as arbitrary unions of sets $U \subseteq X$ such that $U \in \mathcal{N}_x$ for every $x \in U$.^[20] The equivalence follows bidirectionally. Given a topological space $(X, \mathcal{T})$, define $\mathcal{N}_x = \{ U \subseteq X \mid \exists V \in \mathcal{T}, x \in V \subseteq U \}$; this satisfies the axioms because open sets contain their points and are closed under finite intersections and arbitrary unions, with the fourth axiom holding as open sets are neighborhoods of all their points. Conversely, given a neighborhood system $\{\mathcal{N}_x\}_{x \in X}$, the collection $\mathcal{T}$ of unions of "saturated" sets (those containing a neighborhood of each point) forms a topology: $\emptyset, X \in \mathcal{T}$ vacuously; unions preserve saturation; and for finite intersections, if $U_1, \dots, U_n \in \mathcal{T}$, their intersection contains, for each $y$, a neighborhood $M_y \subseteq \bigcap U_i$ by axiom 2 and 3 applied iteratively. A basis for $\mathcal{T}$ consists of sets $B \in \mathcal{N}_x$ for some $x$, ensuring the open sets match. Neighborhoods recover the filter structure where a net $(x_\lambda)$ converges to $x$ if every $N \in \mathcal{N}_x$ contains a tail $\{ x_\lambda \mid \lambda \geq \lambda_0 \}$, mirroring convergence in the induced topology.^[19] Convergence spaces, where convergence of filters or nets to points is axiomatized directly (e.g., constant nets converge, and limits preserve unions), served as precursors to these formulations but yield strict equivalents only when restricted to satisfy the neighborhood filter properties above.^[22]

Continuous Functions

Definition

A function $f: (X, \tau) \to (Y, \sigma)$ between topological spaces $(X, \tau)$ and $(Y, \sigma)$ is continuous if the preimage $f^{-1}(V)$ of every open set $V \in \sigma$ is an open set in $\tau$.^[32] An equivalent characterization is that $f$ is continuous if and only if the preimage $f^{-1}(C)$ of every closed set $C$ in $Y$ is closed in $X$. To see this equivalence, suppose $f$ is continuous in the open set sense. If $C \subseteq Y$ is closed, then its complement $Y \setminus C$ is open, so $f^{-1}(Y \setminus C) = X \setminus f^{-1}(C)$ is open, implying $f^{-1}(C)$ is closed. Conversely, if $f^{-1}(C)$ is closed for every closed $C$, then for any open $V \subseteq Y$, the complement $Y \setminus V$ is closed, so $f^{-1}(Y \setminus V) = X \setminus f^{-1}(V)$ is closed, implying $f^{-1}(V)$ is open.^[32] Another equivalent definition is that $f$ is continuous at every point $x \in X$ if for every neighborhood $N$ of $f(x)$ in $Y$, there exists a neighborhood $M$ of $x$ in $X$ such that $f(M) \subseteq N$. This local condition extends to global continuity on $X$.^[32] The identity function $\mathrm{id}_X: (X, \tau) \to (X, \tau)$ is continuous, as its preimage of any open set $U \in \tau$ is $U$ itself, which is open. Similarly, any constant function $c: (X, \tau) \to (Y, \sigma)$ with $c(x) = y_0$ for all $x \in X$ and fixed $y_0 \in Y$ is continuous, since the preimage of an open set $V \in \sigma$ is either empty (if $y_0 \notin V$) or all of $X$ (if $y_0 \in V$), both of which are open in any topology.^[32]

Basic properties

A fundamental property of continuous functions between topological spaces is that their composition is also continuous. Specifically, if $f: X \to Y$ and $g: Y \to Z$ are continuous functions between topological spaces $X$, $Y$, and $Z$, then the composition $g \circ f: X \to Z$ is continuous.^[32] The converse does not hold: if $g \circ f$ is continuous, it does not necessarily follow that both $f$ and $g$ are continuous. For instance, if $f$ is a constant function (which is continuous) and $g$ is discontinuous, then $g \circ f$ remains constant and hence continuous.^[33] The subspace topology ensures that restrictions of continuous functions remain continuous. If $f: X \to Y$ is continuous and $A \subseteq X$ is equipped with the subspace topology, then the restriction $f|_A: A \to Y$ is continuous.^[34] Moreover, the inclusion map $i: A \to X$, defined by $i(a) = a$ for $a \in A$, is continuous with respect to the subspace topology on $A$.^[34] A homeomorphism is a bijective continuous function whose inverse is also continuous.^[32] This establishes a topological equivalence between spaces, preserving all topological properties. In metric spaces, isometries—distance-preserving bijections—are homeomorphisms, as they are continuous and have continuous inverses.^[35] The pasting lemma provides a method for constructing continuous functions on unions of sets. If $X = A \cup B$ where $A$ and $B$ are either both open or both closed in $X$, and $f: A \to Y$, $g: B \to Y$ are continuous functions that agree on $A \cap B$, then the function $h: X \to Y$ defined by $h|_A = f$ and $h|_B = g$ is continuous.^[36] Continuous functions preserve limits of convergent sequences and, more generally, nets. If $f: X \to Y$ is continuous and a sequence (or net) $(x_\alpha)$ in $X$ converges to $x \in X$, then $f(x_\alpha)$ converges to $f(x)$ in $Y$.^[37]

Examples

Standard topologies

The discrete topology on a set $X$ is defined as the topology $\mathcal{T}$ where every subset of $X$ is an open set, i.e., $\mathcal{T} = \mathcal{P}(X)$, the power set of $X$.^[38] This makes it the finest (largest) topology on $X$, as it contains the maximum number of open sets possible while satisfying the topology axioms.^[39] In this topology, singletons $\{x\}$ for any $x \in X$ are open, and every function from a space with the discrete topology to any other topological space is continuous, since preimages of open sets are always open.^[38] The trivial topology, also known as the indiscrete topology, on a set $X$ consists solely of the empty set $\emptyset$ and $X$ itself as open sets.^[40] It is the coarsest (smallest) topology on $X$, containing the minimal collection of sets required to form a topology.^[41] Any function into a space equipped with the trivial topology is continuous, as the preimage of any open set in the codomain is either $\emptyset$ or the entire domain, both of which are open in any topology.^[40] This topology highlights extremal cases in topological properties, such as universal connectedness, where the space is connected but lacks nontrivial open sets.^[41] The cofinite topology on a set $X$ is generated by declaring a set open if its complement is finite or if it is empty.^[42] Equivalently, the closed sets are the finite subsets of $X$ and $X$ itself.^[43] For infinite $X$, this topology is $T_1$ (points are closed) but not Hausdorff, as any two nonempty open sets must intersect: their complements are finite, so their intersection's complement is a finite union of finites, hence finite, making the intersection cofinite and thus open and nonempty.^[42] It is Hausdorff if and only if $X$ is finite, in which case it coincides with the discrete topology.^[43] The cofinite topology is useful for studying spaces where "large" sets behave like opens, often appearing in examples of non-Hausdorff manifolds or algebraic geometry contexts.^[44] For an uncountable set $X$, the cocountable topology (or countable complement topology) defines a set as open if its complement in $X$ is at most countable or if it is empty.^[45] The closed sets are thus the countable subsets of $X$ and $X$ itself.^[46] Similar to the cofinite case, on uncountable $X$, it is $T_1$ but not Hausdorff, since the intersection of any two nonempty opens has countable complement (union of countables), hence is cocountable and nonempty.^[45] This topology is particularly relevant for uncountable sets, where it distinguishes "small" (countable) closed sets from "large" opens, and on countable $X$, it reduces to the discrete topology.^[46] It serves as an example in convergence theory, where sequences converge if their range avoids the limit point's complement.^[46]

Topologies from metrics

A metric on a set $X$, denoted $(X, d)$, defines a topology known as the metric topology or topology induced by the metric. In this topology, a subset $U \subseteq X$ is open if for every $x \in U$, there exists $\varepsilon > 0$ such that the open ball $B(x, \varepsilon) = \{ y \in X \mid d(x, y) < \varepsilon \}$ is contained in $U$. The collection of all such open balls forms a basis for the topology, meaning every open set is a union of these balls.^[47] A topological space is metrizable if its topology can be induced by some metric on the underlying set. Not all topological spaces are metrizable, but certain conditions guarantee metrizability. The Urysohn metrization theorem states that every second-countable regular Hausdorff space is metrizable, providing a sufficient criterion for the existence of such a metric.^[48][49] Prominent examples of metric-induced topologies include the Euclidean space $\mathbb{R}^n$ equipped with the standard Euclidean metric $d(x, y) = \|x - y\|_2 = \sqrt{\sum_{i=1}^n (x_i - y_i)^2}$, which generates the usual Euclidean topology where open sets are unions of open balls (or, in $\mathbb{R}$, open intervals). Another example is the discrete metric on any set $X$, defined by $d(x, y) = 0$ if $x = y$ and $d(x, y) = 1$ otherwise; this induces the discrete topology, in which every subset of $X$ is open, as open balls of radius less than or equal to 1 are singletons.^[50][51] While many familiar spaces are metrizable, counterexamples exist, such as the long line, a linearly ordered topological space constructed as the ordinal $\omega_1 \times [0,1)$ with the order topology, which is locally Euclidean but non-metrizable because it is sequentially compact yet not compact—a distinction impossible in metrizable spaces.^[52][47]

Topologies from order and algebra

Topologies arising from order relations and algebraic structures provide natural ways to endow sets with topological properties that reflect their underlying order or algebraic operations.

Order Topology

Given a totally ordered set $(X, \leq)$, the order topology is the topology generated by the subbasis consisting of all open rays of the form $(-\infty, b) = \{x \in X \mid x < b\}$ and $(a, \infty) = \{x \in X \mid a < x\}$ for $a, b \in X$.^[53] This subbasis generates a basis of open intervals $(a, b) = \{x \in X \mid a < x < b\}$ where applicable, along with the rays at the boundaries if the order is unbounded. The order topology is Hausdorff if and only if the order is a total order without consecutive elements.^[53] A prominent example is the order topology on the real numbers $\mathbb{R}$ with the standard ordering $\leq$. In this case, the order topology coincides precisely with the standard Euclidean topology on $\mathbb{R}$, as the open intervals $(a, b)$ form a basis for both.^[54] Another example is the Sorgenfrey line, which equips $\mathbb{R}$ with the lower limit topology (or Sorgenfrey topology). Here, a basis consists of half-open intervals of the form $[a, b)$ for $a < b$. This topology is finer than the standard topology on $\mathbb{R}$, meaning every standard open set is Sorgenfrey-open, but not conversely; for instance, the half-open interval $[0, 1)$ is Sorgenfrey-open but not open in the standard topology.^[55] The Sorgenfrey line is hereditarily Lindelöf, normal, and paracompact but fails to be second-countable.^[56]

Algebraic Topologies

Algebraic structures induce topologies that capture geometric or arithmetic properties through their operations or ideals. The Zariski topology on the affine space $\mathbb{A}^n_k$ over a field $k$ defines closed sets as the zero loci $V(I) = \{p \in \mathbb{A}^n_k \mid f(p) = 0 \ \forall f \in I\}$, where $I$ is an ideal in the polynomial ring $k[x_1, \dots, x_n]$.^[57] Open sets are complements of such algebraic varieties. This topology is $T_0$ but rarely Hausdorff unless $n=0$, and it is Noetherian, meaning every descending chain of closed sets stabilizes.^[57] The Zariski topology facilitates the study of algebraic geometry by making polynomial functions continuous.^[58] On the integers $\mathbb{Z}$, the profinite topology (induced by the profinite completion $\hat{\mathbb{Z}} = \varprojlim_{n} \mathbb{Z}/n\mathbb{Z}$) has a basis of neighborhoods of 0 given by the subgroups of finite index, or equivalently, it is the initial topology making all natural projections $\mathbb{Z} \to \mathbb{Z}/n\mathbb{Z}$ (with discrete topology on the quotients) continuous.^[59] In this topology, $\mathbb{Z}$ is dense in $\hat{\mathbb{Z}}$, and arithmetic progressions form a basis for the open sets. The profinite topology on $\mathbb{Z}$ is useful in number theory for studying Galois representations and infinite Galois groups.^[59]

Group Topologies

For a group $G$, a group topology is a topology rendering $G$ a topological group, where the multiplication $G \times G \to G$ and inversion $G \to G$ are continuous maps.^[60] The initial such topology with respect to a family of continuous homomorphisms $\phi_i: G \to H_i$ to known topological groups $H_i$ is the coarsest topology making all $\phi_i$ continuous, thereby ensuring the group operations inherit continuity from the targets.^[60] This construction underlies examples like the profinite topology on $\mathbb{Z}$, where the family consists of projections to finite cyclic groups.^[59]

Topologies from other constructions

The product topology on the Cartesian product $X \times Y$ of two topological spaces $(X, \mathcal{T}_X)$ and $(Y, \mathcal{T}_Y)$ is generated by the basis consisting of all sets of the form $U \times V$, where $U \in \mathcal{T}_X$ and $V \in \mathcal{T}_Y$. This construction ensures that the projection maps $\pi_X: X \times Y \to X$ and $\pi_Y: X \times Y \to Y$ are continuous.^[61] For finite products of more than two spaces, the product topology is defined iteratively by applying this basis construction successively.^[62] A key result concerning infinite products is Tychonoff's theorem, which states that the product of any collection of compact topological spaces, equipped with the product topology, is compact.^[63] The quotient topology arises from an equivalence relation $\sim$ on a topological space $X$, defining a new space $X / \sim$ whose points are the equivalence classes. The quotient map $q: X \to X / \sim$ sends each point to its equivalence class, and a subset $W \subseteq X / \sim$ is open if and only if its preimage $q^{-1}(W)$ is open in $X$.^[64] This topology makes $q$ continuous and is the finest such topology on $X / \sim$.^[65] A classic example is the circle $S^1$, obtained as the quotient of the unit interval $[0,1]$ under the equivalence relation identifying $0$ and $1$, or equivalently as $\mathbb{R} / \mathbb{Z}$, where points differing by integers are identified.^[66] The subspace topology on a subset $Y \subseteq X$ of a topological space $(X, \mathcal{T})$ equips $Y$ with the collection of open sets $\{ Y \cap U \mid U \in \mathcal{T} \}$.^[67] This induced topology ensures that the inclusion map $i: Y \to X$ is continuous and is the coarsest topology on $Y$ making $i$ continuous.^[68] Properties of open sets in the ambient space transfer accordingly: if $Y$ is open in $X$, then open sets in $Y$ coincide with those open in $X$ intersected with $Y$; if $Y$ is closed, a similar relation holds for closed sets.^[69] For a set $X$ and a family of maps $\{f_i: X \to Y_i\}_{i \in I}$ from $X$ to topological spaces $(Y_i, \mathcal{T}_i)$, the initial topology on $X$ is the coarsest topology making all $f_i$ continuous, generated as the topology with subbasis $\{ f_i^{-1}(V_i) \mid V_i \in \mathcal{T}_i, i \in I \}$.^[70] Dually, for a family of maps $\{g_j: Z_j \to X\}_{j \in J}$ into a set $X$, the final topology on $X$ is the finest topology making all $g_j$ continuous, consisting of sets $W \subseteq X$ such that $g_j^{-1}(W)$ is open in $Z_j$ for each $j$.^[71] The product topology is an instance of the initial topology with respect to the projections, while the quotient topology is the final topology with respect to the quotient map.^[70]

Topological Properties

Separation axioms

The separation axioms provide a hierarchy of conditions on topological spaces that ensure varying degrees of distinguishability between points and sets using open neighborhoods or continuous functions. These axioms are fundamental in general topology, as they characterize spaces where geometric intuition from Euclidean spaces holds, such as the ability to separate distinct points without overlap. Weaker axioms focus on point separation, while stronger ones extend to separating closed sets.^[72] A topological space $X$ satisfies the T$_0$ axiom, also known as the Kolmogorov axiom, if for any two distinct points $x, y \in X$, there exists an open set $U \subseteq X$ such that either $x \in U$ and $y \notin U$, or $y \in U$ and $x \notin U$. This minimal separation condition ensures that no two points are topologically indistinguishable, though it does not require symmetric separation. The axiom is named after Andrey Kolmogorov, based on his unpublished manuscript.^[73][74] The T$_1$ axiom, or Fréchet axiom, strengthens T$_0$ by requiring that every singleton set $\{x\}$ for $x \in X$ is closed. Equivalently, for any two distinct points $x, y \in X$, there exist open sets $U$ containing $x$ but not $y$, and $V$ containing $y$ but not $x$. This implies that finite sets are closed, as the complement of a finite set is a union of open singletons' complements. The axiom is attributed to Maurice Fréchet, who discussed related ideas in his 1906 thesis on functional calculus, laying groundwork for abstract spaces where points can be isolated by opens.^[75][74] A space $X$ is T$_2$, or Hausdorff, if for every pair of distinct points $x, y \in X$, there exist disjoint open neighborhoods $U \ni x$ and $V \ni y$. This symmetric separation prevents points from being "stuck together" and is crucial for many theorems, such as the uniqueness of limits. The axiom originates from Felix Hausdorff's 1914 monograph Grundzüge der Mengenlehre, where it appears as Axiom II, $\mathfrak{o}_2$, in the context of neighborhood systems for general topological spaces.^[74] Beyond point separation, regularity addresses separation from closed sets. A space $X$ is regular if for every point $x \in X$ and every closed set $C \subseteq X$ with $x \notin C$, there exist disjoint open sets $U \ni x$ and $V \supseteq C$. The T$_3$ axiom combines regularity with the T$_1$ condition. This allows points to be separated from arbitrary closed sets, enhancing local structure. The concept evolved in the 1920s through works on metrizability, with formalization in texts building on Hausdorff's framework.^[76][74] A refinement is complete regularity, where for every $x \in X$ and closed $C \subseteq X$ with $x \notin C$, there exists a continuous function $f: X \to [0,1]$ such that $f(x) = 0$ and $f(C) = 1$. A completely regular T$_1$ space is called T$_{3.5}$ or Tychonoff. This functional separation enables embedding into products of intervals and is key for compactifications. The notion is due to Andrey Tychonoff, who developed it in 1930 to study topological extensions and products.^[77][78] The strongest common axiom is T$_4$, or normality: for any two disjoint closed sets $A, B \subseteq X$, there exist disjoint open sets $U \supseteq A$ and $V \supseteq B$. A normal T$_1$ space is T$_4$. This global separation supports theorems like Urysohn's lemma, which constructs continuous functions separating such sets. The axiom traces to Pavel Urysohn's 1922 paper on metrication of compact spaces, where he proved equivalent functional forms for normal spaces.^[79][77] These axioms form an implication chain: every T$_4$ space is T$_3$, every T$_3$ space is T$_2$, every T$_2$ space is T$_1$, and every T$_1$ space is T$_0$. Without T$_1$, the hierarchy branches, as regularity or normality may hold without point separation (e.g., R$_0$ symmetric spaces). For instance, the real line with standard topology satisfies all, while the cofinite topology on an infinite set is T$_1$ but not T$_2$. These relations ensure that stronger axioms preserve desirable properties like metrizability in Euclidean-like spaces.^[74][72]

Connectedness and compactness

A topological space $X$ is defined as connected if it cannot be expressed as the union of two nonempty disjoint open subsets. This property captures the intuitive notion that the space is "in one piece" and cannot be separated into distinct open components. Equivalently, the only subsets of $X$ that are both open and closed are the empty set and $X$ itself.^[80] Path-connectedness provides a stronger form of connectedness, where for any two points $x, y \in X$, there exists a continuous path $\gamma: [0,1] \to X$ such that $\gamma(0) = x$ and $\gamma(1) = y$. Every path-connected space is connected, but the converse does not hold; for instance, the topologist's sine curve is connected but not path-connected. The real line $\mathbb{R}$ with the standard topology is connected, as any nonempty open subset is a union of open intervals, and separating it into disjoint nonempty opens would contradict the intermediate value property of intervals.^[81] In contrast, the rational numbers $\mathbb{Q}$ as a subspace of $\mathbb{R}$ are disconnected, since they can be partitioned into opens like $\mathbb{Q} \cap (-\infty, \sqrt{2})$ and $\mathbb{Q} \cap (\sqrt{2}, \infty)$, both nonempty and disjoint.^[82] Continuous images preserve connectedness: if $f: X \to Y$ is continuous and $X$ is connected, then $f(X)$ is connected in $Y$.^[83] Continuous functions also map compact sets to compact sets. A topological space $X$ is compact if every open cover of $X$ admits a finite subcover. In Hausdorff spaces, compactness is equivalent to limit point compactness, where every infinite subset has a limit point. Local compactness strengthens this locally: every point in $X$ has a neighborhood basis consisting of compact subsets. The Heine-Borel theorem states that in $\mathbb{R}^n$ with the Euclidean topology, a subset is compact if and only if it is closed and bounded. Tychonoff's theorem extends compactness to products: the product of any collection of compact spaces is compact in the product topology.

Countability axioms

Countability axioms in topology concern properties that impose restrictions on the "size" of certain topological structures, specifically requiring countable collections to suffice for covering or generating the space. These axioms are crucial for ensuring that abstract topological spaces behave similarly to more concrete ones, such as metric spaces, where countable structures often play a central role in defining limits, continuity, and density.^[84] A topological space $X$ is first-countable if, for each point $x \in X$, there exists a countable local basis at $x$, that is, a countable family $\{B_n(x)\}_{n \in \mathbb{N}}$ of open neighborhoods of $x$ such that every open neighborhood $U$ of $x$ contains some $B_n(x)$.^[85] In first-countable spaces, a function is continuous at a point if and only if it is sequentially continuous at that point, meaning that whenever a sequence converges to the point, the image sequence converges to the image point; this equivalence simplifies the study of continuity by reducing it to sequential criteria.^[86] A topological space $X$ is second-countable if its topology has a countable basis, that is, a countable collection $\mathcal{B} = \{B_n\}_{n \in \mathbb{N}}$ of open sets such that every open set in $X$ is a union of members of $\mathcal{B}$.^[84] Second-countable spaces are Lindelöf, meaning every open cover of $X$ admits a countable subcover, which follows from selecting countably many basis elements to refine the cover.^[85] Moreover, second-countable spaces are separable, possessing a countable dense subset $D \subseteq X$ such that the closure of $D$ is all of $X$; to construct such a $D$, choose one point from each nonempty basis element using the axiom of countable choice.^[84] A topological space $X$ is separable if it contains a countable dense subset, as defined above.^[87] Separability ensures that the space can be "approximated" by a countable set in terms of closure points. Separability implies the countable chain condition (ccc), which states that every family of pairwise disjoint nonempty open sets in $X$ is at most countable; this follows because each such open set must intersect the dense subset in distinct points, limiting the family to countable size.^[88] The countable chain condition (ccc) requires that no uncountable collection of pairwise disjoint nonempty open sets exists in $X$.^[89] This axiom prevents the topology from having "too many" independent open structures and is particularly relevant in spaces where separability does not fully capture countability constraints. Among these axioms, second-countability implies first-countability, since a countable global basis restricts to a countable local basis at each point by taking intersections with neighborhoods.^[90] Second-countability also implies separability and the ccc in general topological spaces.^[84] In metric spaces, second-countability, separability, and the Lindelöf property are equivalent, while all metrizable spaces are first-countable.^[85]

Other classifications

A topological space $X$ is paracompact if every open cover of $X$ admits a locally finite open refinement, and $X$ is Hausdorff.^[91] This property ensures that paracompact Hausdorff spaces support partitions of unity subordinate to any open cover, facilitating the extension of continuous functions and the study of differential structures on manifolds.^[92] For instance, every metric space and every compact Hausdorff space is paracompact.^[93] Dimension theory in topology provides invariants that measure the "dimensionality" of a space through inductive and covering dimensions. The small inductive dimension, denoted $\operatorname{ind} X$, is defined recursively: $\operatorname{ind} \emptyset = -1$, and for a nonempty space $X$, $\operatorname{ind} X = n+1$ if every pair of disjoint closed sets can be separated by disjoint open sets whose boundaries have $\operatorname{ind} \leq n$, while the large inductive dimension, $\operatorname{Ind} X$, uses closures instead of boundaries in the separation condition.^[94] The covering dimension, $\dim X$, is the minimal integer $n$ such that every finite open cover has an open refinement where no point lies in more than $n+1$ sets.^[95] In Euclidean space $\mathbb{R}^n$, all three dimensions equal $n$.^[96] Cardinal invariants quantify the "size" of a topological space in terms of bases and dense subsets. The weight $w(X)$ is the minimal cardinality of a base for the topology of $X$, representing the smallest basis size needed to generate all open sets.^[97] The density $d(X)$ is the minimal cardinality of a dense subset of $X$, indicating the smallest set whose closure is the entire space.^[97] For example, in a separable metric space, $d(X) = \aleph_0$, while $w(X)$ can be larger, as in the Hilbert cube where $w(X) = \mathfrak{c}$.^[98] A topological space is uniformizable if its topology arises from a compatible uniformity, a structure generalizing metrics to define uniform continuity without distances.^[99] Completely regular spaces are precisely the uniformizable ones, and all regular Hausdorff ($T_3$) metrizable spaces admit such uniform structures, enabling the uniform extension of continuous functions.^[100] The Baire category theorem characterizes certain spaces where "residual" sets—countable intersections of dense open sets—are dense. A space $X$ is a Baire space if the intersection of any countable collection of dense open subsets is dense in $X$.^[101] Complete metric spaces and locally compact Hausdorff spaces satisfy this property, implying that meager sets (countable unions of nowhere dense sets) have empty interior.^[102] This framework is crucial for proving the existence of generic points in function spaces.^[103]