Open covers are the language topology uses for turning local information into global information. A single [open set](/page/Open%20Set) records a region in which nearby points are available; a family of open sets records many such local regions at once. When those regions cover a whole space, or a chosen subset of it, they let us ask whether the object can be understood by finitely many local pieces, by smaller basic pieces, or by pieces arranged with controlled overlap. This is why open covers appear in [compactness](/page/Compact%20Space), [continuity](/page/Continuity), manifold theory, sheaf theory, and partitions of unity. The ordinary word "cover" is too broad for topology. A set can be covered by arbitrary subsets that ignore the topology, but local arguments depend on openness: if a point lies in an open member of the cover, then a whole neighbourhood of that point lies in the same member. Open covers are designed to preserve that neighbourhood information. ## Definition The basic object is a family of open subsets whose union is the space under discussion. The index set is allowed to be finite, countably infinite, or uncountable; the topology controls the members of the family, not the size of the family. [definition: Open Cover] Let $(X, \tau)$ be a [topological space](/page/Topological%20Space). An open cover of $X$ is a collection $\mathcal U \subset \tau$ such that \begin{align*} X &= \bigcup_{U \in \mathcal U} U. \end{align*} [/definition] Thus an open cover is not just any family whose union is $X$; each member must be open in the topology $\tau$. The condition says that every point $x \in X$ belongs to at least one member $U \in \mathcal U$. Many arguments, however, do not need to cover the entire ambient space. They focus on a subset whose points must be reached by ambient open neighbourhoods. [definition: Open Cover of a Subset] Let $(X, \tau)$ be a topological space, and let $A \subset X$. An open cover of $A$ in $X$ is a collection $\mathcal U \subset \tau$ such that \begin{align*} A &\subset \bigcup_{U \in \mathcal U} U. \end{align*} [/definition] If $A=X$, this reduces to the preceding definition. The phrase "in $X$" matters: if $A$ has the [subspace topology](/page/Subspace%20Topology), some sets may be open in $A$ without being open in $X$. Once a cover is available, the next question is whether all its members are actually needed, which leads to the language of subcovers. [definition: Subcover] Let $\mathcal U$ be a cover of a set $A$. A subcover of $\mathcal U$ is a subcollection $\mathcal V \subset \mathcal U$ such that \begin{align*} A &\subset \bigcup_{V \in \mathcal V} V. \end{align*} [/definition] Subcovers only delete members from a cover, so they measure economy rather than precision. Many local arguments require a different operation: replace coarse neighbourhoods by smaller ones that still sit inside the original cover. That need is captured by refinement. [definition: Refinement of a Cover] Let $\mathcal U$ and $\mathcal V$ be covers of a set $A \subset X$. The cover $\mathcal V$ refines $\mathcal U$ if for every $V \in \mathcal V$ there exists $U \in \mathcal U$ such that \begin{align*} V &\subset U. \end{align*} [/definition] To construct functions by summing local contributions, topology must prevent infinitely many cover elements from interacting near a single point. The cover condition alone does not provide this control. Local finiteness is the standard hypothesis that makes such local sums and choices behave well. [definition: Locally Finite Open Cover] Let $(X, \tau)$ be a topological space. A locally finite open cover of $X$ is an open cover $\mathcal U$ of $X$ such that for every $x \in X$ there exists an open set $W \in \tau$ with $x \in W$ and \begin{align*} \{U \in \mathcal U : U \cap W \neq \varnothing\} \end{align*} is finite. [/definition] Local finiteness is stronger than saying each point lies in finitely many members. It asks for a neighbourhood on which only finitely many members of the cover can appear. ## Equivalent Characterisations The definition of open cover can be read pointwise. This viewpoint is useful when constructing covers, because it turns a union condition into a local membership condition. [quotetheorem:8483] This statement is often the most convenient way to verify that a family covers. It also explains why covers behave like local data: each point is assigned at least one open neighbourhood from the family. The same pointwise test should adapt to subsets, but now the quantifier is restricted to the subset being covered. [quotetheorem:8484] This form is especially common in metric and analytic arguments, where a neighbourhood is chosen separately around each point of a set. If those choices need to be tracked by parameters, it is helpful to record the cover as an indexed family rather than as an unindexed collection. [definition: Indexed Open Cover] Let $(X, \tau)$ be a topological space, and let $I$ be an index set. An indexed open cover of $X$ is a family $(U_i)_{i \in I}$ with $U_i \in \tau$ for every $i \in I$ and \begin{align*} X &= \bigcup_{i \in I} U_i. \end{align*} [/definition] The indexed and unindexed viewpoints carry the same covering condition, but they remember different information. An indexed family can repeat the same open set under different indices, while a collection forgets repetitions. ## Standard Examples The real line has many open covers that cannot be reduced to finitely many sets. This is the first place where open covers reveal a global property rather than a local one. [example: Expanding Intervals Cover the Real Line] In $\mathbb R$ with its usual topology, consider \begin{align*} \mathcal U = \{(-n,n) : n \in \mathbb N\}. \end{align*} Each member $(-n,n)$ is open in the usual topology, so it remains to check that the union is all of $\mathbb R$. Let $x \in \mathbb R$. By the [Archimedean property](/theorems/737), choose $n \in \mathbb N$ with $n>|x|$. Then $-|x| \le x \le |x|$, and since $|x|<n$, we have $-n<x<n$. Hence $x \in (-n,n)$, so every $x\in\mathbb R$ lies in at least one member of $\mathcal U$. No finite subcollection covers $\mathbb R$. Suppose a finite subcollection is chosen: \begin{align*} \{(-n_1,n_1),\ldots,(-n_k,n_k)\}. \end{align*} Let \begin{align*} N=\max\{n_1,\ldots,n_k\}. \end{align*} For each $i$, $n_i\le N$, so $(-n_i,n_i)\subset (-N,N)$. Therefore \begin{align*} \bigcup_{i=1}^k (-n_i,n_i) \subset (-N,N). \end{align*} Also $N=n_j$ for some $j$, so $(-N,N)=(-n_j,n_j)$ is one of the chosen intervals, and hence \begin{align*} (-N,N)\subset \bigcup_{i=1}^k (-n_i,n_i). \end{align*} Thus the finite union is exactly $(-N,N)$. The real number $N$ itself is not in $(-N,N)$, since membership would require $N<N$, which is false. Hence this finite subcollection does not cover $\mathbb R$. The cover is therefore genuinely infinite: it covers every real number, but no finite part of it does. [/example] This example shows why compactness cannot mean "has open covers"; every topological space has at least the cover by itself. Compactness instead asks that every open cover can be reduced to finitely many members. For a bounded closed interval, the same style of cover behaves differently. The geometry of the interval forces finitely many sufficiently large members to cover it. [example: A Finite Subcover of a Compact Interval] In $\mathbb R$ with its usual topology, let \begin{align*} \mathcal U = \{(-n,n) : n \in \mathbb N\}. \end{align*} Each set $(-n,n)$ is open in the usual topology on $\mathbb R$, so every member of $\mathcal U$ is open. We show that the single member $(-3,3)$ already covers $[-2,2]$. Let $x \in [-2,2]$. By membership in the closed interval, $-2 \le x \le 2$. Since $-3<-2$ and $2<3$, the inequalities combine to give \begin{align*} -3 < -2 \le x \le 2 < 3. \end{align*} Hence $-3<x<3$, so $x\in(-3,3)$. Therefore \begin{align*} [-2,2] \subset (-3,3). \end{align*} Because $3\in\mathbb N$, the interval $(-3,3)$ is a member of $\mathcal U$, and the subcollection \begin{align*} \{(-3,3)\} \subset \mathcal U \end{align*} has exactly one element. Thus $\{(-3,3)\}$ is a finite subcover of $[-2,2]$ in $\mathbb R$. This shows that the same expanding-interval cover that cannot be reduced finitely on all of $\mathbb R$ becomes finite when restricted to the bounded interval $[-2,2]$. [/example] The contrast with the previous example is the seed of the open-cover definition of compactness. The same cover fails to have a finite subcover for all of $\mathbb R$ but has one for $[-2,2]$. The openness condition is not decorative. A family may cover every point and still fail to be an open cover because its members are not open. [example: A Cover That Is Not an Open Cover] In $\mathbb R$ with the usual topology, consider \begin{align*} \mathcal C = \{[n,n+1] : n \in \mathbb Z\}. \end{align*} This family covers $\mathbb R$. Indeed, let $x\in\mathbb R$. By the defining property of the floor function, there is an integer $n=\lfloor x\rfloor$ such that \begin{align*} n \le x < n+1. \end{align*} Since $x<n+1$ implies $x\le n+1$, we have \begin{align*} n \le x \le n+1. \end{align*} Therefore $x\in[n,n+1]$, and because $n\in\mathbb Z$, the interval $[n,n+1]$ is a member of $\mathcal C$. However, $\mathcal C$ is not an open cover of $\mathbb R$. For any $n\in\mathbb Z$, the left endpoint $n$ belongs to $[n,n+1]$. If $\varepsilon>0$, then \begin{align*} n-\frac{\varepsilon}{2}<n<n+\frac{\varepsilon}{2}, \end{align*} so $n-\frac{\varepsilon}{2}\in(n-\varepsilon,n+\varepsilon)$. But \begin{align*} n-\frac{\varepsilon}{2}<n, \end{align*} so $n-\frac{\varepsilon}{2}\notin[n,n+1]$. Thus no open interval around $n$ is contained in $[n,n+1]$, so $[n,n+1]$ is not open in the usual topology. Since every member of an open cover must be open, $\mathcal C$ is a cover of $\mathbb R$ but not an open cover. [/example] This example separates the set-theoretic idea of covering from the topological idea of open covering. The latter is the one compatible with local continuity and neighbourhood arguments. Refinements let us replace a cover by smaller, better adapted pieces. In metric spaces, balls give a canonical source of such refinements. [example: Ball Refinement of an Open Cover] Let $(X,d)$ be a [metric space](/page/Metric%20Space), and let $\mathcal U$ be an open cover of $X$. For each $x\in X$, the covering condition gives at least one set $U_x\in\mathcal U$ such that $x\in U_x$. Since $U_x$ is open in the metric topology, there exists $r_x>0$ such that \begin{align*} B(x,r_x)\subset U_x. \end{align*} Define \begin{align*} \mathcal V=\{B(x,r_x):x\in X\}. \end{align*} Each member of $\mathcal V$ is open. Indeed, fix $x\in X$ and let $y\in B(x,r_x)$. Then $d(x,y)<r_x$, so \begin{align*} \varepsilon=r_x-d(x,y)>0. \end{align*} If $z\in B(y,\varepsilon)$, then $d(y,z)<\varepsilon$, and the triangle inequality gives \begin{align*} d(x,z)\le d(x,y)+d(y,z)<d(x,y)+\varepsilon. \end{align*} Substituting $\varepsilon=r_x-d(x,y)$ gives \begin{align*} d(x,y)+\varepsilon=d(x,y)+r_x-d(x,y)=r_x. \end{align*} Thus $d(x,z)<r_x$, so $z\in B(x,r_x)$. Hence $B(y,\varepsilon)\subset B(x,r_x)$, proving that $B(x,r_x)$ is open. The family $\mathcal V$ covers $X$ because, for every $x\in X$, the point $x$ lies in its own ball: \begin{align*} d(x,x)=0<r_x. \end{align*} Therefore $x\in B(x,r_x)\in\mathcal V$. Finally, $\mathcal V$ refines $\mathcal U$ because each member $B(x,r_x)$ was chosen with \begin{align*} B(x,r_x)\subset U_x \end{align*} for some $U_x\in\mathcal U$. Thus the arbitrary open cover $\mathcal U$ has been replaced by an open cover by metric balls, with every ball still sitting inside one of the original cover elements. [/example] This construction is a recurring move in analysis: replace arbitrary open sets by metric balls, then use radii to quantify local estimates. ## Properties The most famous use of open covers is the topological definition of compactness. It turns the intuitive idea of a space being globally small into a statement about reducing arbitrary local data to finite local data. [definition: Compact Space] A topological space $(X,\tau)$ is compact if every open cover of $X$ has a finite subcover. [/definition] This definition is one of the reasons open covers are central. It does not mention distances, boundedness, or sequences, so it applies in every topological space. Analysis often needs the same finite-subcover principle for a subset sitting inside a larger space, where the covering sets are open in the ambient topology. [definition: Compact Subset] Let $(X,\tau)$ be a topological space. A subset $K \subset X$ is compact in $X$ if every open cover of $K$ in $X$ has a finite subcover. [/definition] The ambient-space formulation is often more convenient when the covering sets come from $X$, while the intrinsic formulation treats $K$ as a space in its own right. Since compactness is meant to be a property of the subset rather than an accident of notation, these two formulations must agree. The next theorem records that agreement by translating between ambient open sets in $X$ and open sets in the subspace topology on $K$. [quotetheorem:8485] This theorem is useful because it lets arguments move freely between covers by open subsets of $X$ and covers by open subsets of $K$. A key test of compactness is whether it is inherited by natural subspaces. Closed subsets are the most important case, because their complements provide the missing open set needed to cover the ambient compact space. [quotetheorem:8487] This theorem is often used to generate compact examples. In $\mathbb R^n$, it is one half of the Heine--Borel picture: closed bounded sets are compact. The next basic stability question asks what happens when a compact space is mapped continuously into another space. [quotetheorem:8489] The theorem is a compact way to remember the definition of continuity: preimages of open sets are open, and preimages commute with unions. It also supplies the main mechanism behind compactness of continuous images, because finite subcovers can be selected after pulling the cover back to the compact domain. [quotetheorem:305] Open covers are the mechanism behind the result. An open cover of $f(X)$ pulls back to an open cover of $X$; compactness selects finitely many pulled-back sets, and those correspond to finitely many original sets covering $f(X)$. Another common way to make covers usable is to replace arbitrary open sets by basis elements. [quotetheorem:8490] In metric spaces, applying this theorem with the ball basis says arbitrary open covers can be refined by open balls. In manifolds, applying it with a basis of coordinate domains gives refinements by chart domains. For constructions involving functions assigned to cover elements, refinement is not enough; one also needs control over how many sets meet near each point. [quotetheorem:8492] This is the cover-theoretic reason locally finite covers appear in the construction of [partitions of unity](/page/Partition%20of%20Unity). The topology controls not only where functions live, but also how many of them interact near a point. ## Relationship to Other Concepts Open covers are inseparable from compactness. The definition of a [compact space](/page/Compact%20Space) says that every open cover has a finite subcover; many compactness theorems are strategies for extracting such finite subcovers from apparently infinite local data. They also encode continuity in a global way. If $f:X\to Y$ is continuous, every open cover of $Y$ pulls back to an open cover of $X$. Conversely, many continuity tests can be phrased in terms of how covers of the codomain behave under preimage. In metric spaces, open covers are often refined by balls, which connects the topological language to estimates involving radii. This bridge is important in analysis: local boundedness, [uniform continuity](/page/Uniform%20Continuity), Lebesgue numbers, and compactness arguments all move between open covers and metric balls. In manifold theory, open covers organize coordinate charts. A smooth manifold is studied by choosing open sets on which coordinates are available, proving local statements there, and then checking compatibility on overlaps. Refinements by smaller chart domains are common when constructing smooth functions, differential forms, or partitions of unity. In sheaf theory, open covers are part of the gluing mechanism. Local sections over members of a cover can be compared on pairwise intersections, and compatible local data may glue to global data. Thus the cover is not only a collection of sets; it is the scaffold on which local-to-global reasoning is built. ## References James R. Munkres, *Topology* (2000). John L. Kelley, *General Topology* (1955). Walter Rudin, *Principles of Mathematical Analysis* (1976). [Open Set](/page/Open%20Set). [Topological Space](/page/Topological%20Space). [Compact Space](/page/Compact%20Space).