Topological Spaces, Bases, and Subbases

A topological space is a set equipped with a collection of "open" subsets closed under arbitrary unions and finite intersections; this minimal structure axiomatizes continuity, convergence, and "nearness" without metric or distance. A basis is a generating collection of open sets from which every open set arises as a union; a subbasis is an even smaller generating collection whose finite intersections form a basis. Used everywhere in modern mathematics — analysis, geometry, algebraic topology. Codified by Felix Hausdorff (1914) building on Fréchet, Riesz, and Cantor.

A topology on a set $X$ is a family $\tau$ of subsets — the open sets — satisfying three axioms: $\varnothing$ and $X$ are open, arbitrary unions of open sets are open, and finite intersections of open sets are open. This deceptively simple structure is the most general setting in which one can speak of continuity, limits, and connectedness, generalizing the open-interval topology of $\mathbb{R}$. Bases and subbases solve a practical problem: explicitly listing every open set is usually impossible, so we specify a small generating family instead, exactly as a vector-space basis generates all vectors. The surprising fact is just how little structure is required: there is no notion of distance, angle, or size anywhere in the axioms, yet they suffice to develop nearly all of point-set topology.

The Core Insight

Topology abstracts the single feature of distance that actually matters for continuity: which sets are "open" (contain a neighborhood around each of their points). Bases let us generate the entire topology from manageable building blocks, and subbases push this further — you can specify a topology by an arbitrary collection of sets and let finite intersections and unions do the rest. Open sets are the primitive; everything else (closure, continuity, convergence) is defined in terms of them.

Formal Statement

Topological Space

Let $X$ be a set. A topology on $X$ is a collection $\tau \subseteq \mathcal{P}(X)$ of subsets of $X$ such that:

The pair $(X,\tau )$ is a topological space; members of $\tau$ are open sets, and their complements are closed sets. The symbol $\mathcal{P}(X)$ denotes the power set of $X$, $I$ is an arbitrary index set, and $n\in \mathbb{N}$.

Basis for a Topology

A collection $\mathcal{B}\subseteq \mathcal{P}(X)$ is a basis for a topology on $X$ if:

The topology generated by $\mathcal{B}$ consists of all unions of basis elements:

Subbasis

A collection $\mathcal{S}\subseteq \mathcal{P}(X)$ is a subbasis for a topology on $X$ if ${\bigcup }_{S\in \mathcal{S}}S=X$. The topology generated by $\mathcal{S}$ is the topology whose basis is the collection of all finite intersections of members of $\mathcal{S}$:

Generated Topology is Well-Defined

If $\mathcal{B}$ satisfies [4]–[5], then ${\tau }_{\mathcal{B}}$ defined in [6] is a genuine topology on $X$, and $\mathcal{B}\subseteq {\tau }_{\mathcal{B}}$. Moreover, for any collection $\mathcal{S}$ covering $X$, ${\tau }_{\mathcal{S}}$ is the coarsest (smallest) topology containing $\mathcal{S}$:

Why It Holds

That $\tau_{\mathcal{B}}$ Satisfies the Axioms

We verify [1]–[3] for ${\tau }_{\mathcal{B}}$ as defined by [6].

That Finite Intersections of a Subbasis Form a Basis

Let $\mathcal{S}$ cover $X$ and ${\mathcal{B}}_{\mathcal{S}}$ be as in [7].

Therefore ${\mathcal{B}}_{\mathcal{S}}$ is a basis, and ${\tau }_{\mathcal{S}}={\tau }_{{\mathcal{B}}_{\mathcal{S}}}$ is a topology by the previous proof. $■$

Felix Hausdorff | 1868–1942

Hausdorff gave point-set topology its modern axiomatic form in Grundzüge der Mengenlehre (1914). His original definition used neighborhood systems and included what we now call the Hausdorff (T₂) separation axiom — the requirement that distinct points have disjoint neighborhoods. The stripped-down "open sets only" axiomatization in current use was refined by Kuratowski, Alexandroff, and others through the 1920s. A polymath who also worked in set theory (the Hausdorff maximal principle) and wrote literature under the pen name Paul Mongré, Hausdorff and his wife took their own lives in 1942 to escape deportation to a Nazi concentration camp.

Building Intuition

Think of a topology as a recipe for "closeness" that throws away numbers. In a metric space, "$x$ is near $A$" means there are points of $A$ arbitrarily close. Topology keeps only the qualitative residue: a set $U$ is open if every point sits "comfortably inside" $U$ with some wiggle room (a neighborhood) entirely contained in $U$.

A basis is your toolkit of standard wiggle rooms. In $\mathbb{R}$ you don't list all open sets — you declare the open intervals $(a,b)$ as the basis, and every open set is a union of them. The two basis axioms say: (1) every point has at least one tool around it, and (2) where two tools overlap, you can find a smaller tool fitting inside the overlap — so you never get "stuck" trying to express an intersection.

A subbasis is even lazier: you throw down any sets you like (as long as they cover $X$), and the machinery automatically takes finite intersections (to make a basis) then arbitrary unions (to make the topology). It is the topological analog of generating a group from a set of generators — finite words in the generators give everything reachable.

Generating open sets: subbasis → basis → topology in the plane

Worked Examples

Standard Topology on $\mathbb{R}$ from interval basis

Subbasis generating the standard topology on $\mathbb{R}$

Common Misconceptions

A collection that covers $X$ but is NOT a basis

Let $X=\{1,2,3\}$ and $\mathcal{C}=\{\{1,2\},\{2,3\}\}$. This covers $X$ ($\{1,2\}\cup \{2,3\}=X$). But it is not a basis: take $B_1=\{1,2\}$, $B_2=\{2,3\}$ and the point $x=2\in B_1\cap B_2=\{2\}$. There is no member of $\mathcal{C}$ contained in $\{2\}$ that contains $2$, so [5] fails.

However $\mathcal{C}$ is a perfectly good subbasis: forming finite intersections adds $\{2\}=\{1,2\}\cap \{2,3\}$, yielding the basis $\{\{1,2\},\{2,3\},\{2\}\}$, and the generated topology is

Connections

Topological spaces sit at the root of an enormous network. The standard topology on ${\mathbb{R}}^n$ specializes to Metric Spaces, where the basis is the collection of open balls $B(x,r)$ and the metric supplies axiom [5] via the triangle inequality. Bases are the precise tool for defining the Product Topology (where a subbasis of "cylinder sets" ${\pi }_i^{-1}(U)$ is natural) and the Subspace Topology (intersect open sets with the subset). The notion of generating a topology from a subbasis parallels generating a Sigma-Algebra in measure theory and generating a Group from Generators in algebra. Continuity itself — see Continuous Functions Between Topological Spaces — is testable on a basis or even a subbasis alone: $f$ is continuous iff preimages of subbasis elements are open, a major practical economy.

Real-World Applications

1. Functional analysis — the weak and weak-\* topologies on Banach spaces are defined by subbases of the form $\{x:|\varphi (x)-\varphi (x_0)|<\epsilon \}$; this is essential to the Banach–Alaoglu theorem and to PDE solution spaces.
2. Topological data analysis (TDA) — persistent homology builds simplicial complexes from open covers; the Vietoris–Rips and Čech constructions rely on bases of balls to detect "shape" in high-dimensional point clouds (used in genomics, neuroscience).
3. Domain theory in computer science — the Scott topology on a partial order, generated by Scott-open sets as a subbasis, models the semantics of programming languages and lazy evaluation.
4. Digital topology and image processing — the Khalimsky topology on ${\mathbb{Z}}^n$ gives pixels a topological structure so that connectedness and boundaries are well-defined for edge detection.
5. Quantum logic and physics — the Zariski topology (subbasis: complements of vanishing sets of polynomials) underlies algebraic geometry, used in string theory compactifications and coding theory.
6. Sensor networks — coverage problems use the nerve of an open cover (Čech complex); topology certifies that a region is fully monitored with no holes.
7. Optimization & economics — the order topology and product topologies (Tychonoff's theorem) guarantee existence of equilibria and fixed points in infinite-dimensional strategy spaces.

Historical Development

• 1872 — Georg Cantor studies derived sets and limit points of subsets of $\mathbb{R}$, isolating the notion of accumulation point.
• 1893 — Camille Jordan and others develop rigorous notions of open and closed regions for curve theory (Jordan curve theorem).
• 1906 — Maurice Fréchet introduces metric spaces and abstract convergence in his thesis, the first general framework for "limit" beyond ${\mathbb{R}}^n$.
• 1908 — Frigyes Riesz proposes an axiomatic theory of "condensation points," anticipating neighborhood structures.
• 1914 — Felix Hausdorff publishes Grundzüge der Mengenlehre, defining topological spaces via neighborhood axioms (including T₂ separation) — the birth of point-set topology.
• 1922 — Kazimierz Kuratowski gives the equivalent closure-operator axiomatization, clarifying the duality of open and closed sets.
• 1923–1927 — Pavel Alexandroff and Pavel Urysohn develop compactness, the open-cover definition, and metrization theory, cementing the open-set formulation now standard.
• 1929 — Andrey Tychonoff proves the product of compact spaces is compact, showcasing the power of subbases (cylinder sets) in infinite products.
• 1937 — Bourbaki's Topologie générale systematizes filters, bases, and subbases into the canonical modern presentation.

Subtleties & Limits

The open-set axioms are too permissive for many purposes, which is why topologists add separation axioms (T₀ through T₄), countability axioms, and covering axioms. Several deep questions live here:

• Metrization. Which topological spaces arise from a metric? The Nagata–Smirnov and Bing metrization theorems give characterizations via $\sigma$-locally-finite bases, but determining "good" bases for a given space remains delicate. The normal Moore space conjecture (every normal Moore space is metrizable) turned out to be independent of ZFC — its truth depends on set-theoretic axioms like the continuum hypothesis or large cardinals.
• Cardinal invariants. The minimum cardinality of a basis (the weight $w(X)$) and of a local basis (the character $\chi (X)$) are subtle and connect to set theory; questions about them (e.g., the existence of spaces with prescribed invariants) drive ongoing research in set-theoretic topology.
• Pointless topology. Locale theory studies the lattice of open sets directly, without points, and behaves better than classical topology in constructive/intuitionistic settings — yet the exact relationship to point-set spaces (sobriety, spatiality) remains an active area.

See Also