Open cover
An open cover of a topological space (X) is a family (\mathcal U) of open sets whose union contains (X). Since every member of (\mathcal U) is a subset of (X), this condition is ordinarily written as
[ X=\bigcup_{U\in\mathcal U}U. ]
Open covers encode global properties of a space through collections of local neighborhoods. Their principal uses occur in the definitions of compactness, paracompactness, covering dimension, and several forms of local-to-global construction. They also provide the indexing framework for partitions of unity, Čech cohomology, and the gluing operations associated with sheaves.
The term “open” refers to the topology of the space being covered, rather than to a geometric property of the individual subsets. Consequently, the same family may be an open cover for one topology and merely a cover for another.
Definition and notation
A cover of a set (X) is a family (\mathcal A\subseteq\mathcal P(X)) satisfying
[ X\subseteq\bigcup_{A\in\mathcal A}A. ]
When every (A\in\mathcal A) is already a subset of (X), the inclusion is equivalent to equality. An open cover is a cover whose members belong to the topology (\tau) on (X).
Covers are frequently represented as indexed families
[ \mathcal U=(U_i){i\in I}, \qquad X=\bigcup{i\in I}U_i, ]
because different indices may correspond to the same open subset. This distinction is relevant when the index set carries additional information or when constructions assign separate data to coincident members of the cover.
The family ({X}) is an open cover of every topological space. At the opposite extreme, the family of all open subsets of (X) is also an open cover when (X) is nonempty. In a discrete space, every ordinary cover is open because every subset is open. In an indiscrete space, every nonempty open cover necessarily contains the entire space.
For a subset (A\subseteq X), a family of open subsets of (X) covers (A) when
[ A\subseteq\bigcup_{U\in\mathcal U}U. ]
The members of such a family need not be subsets of (A). Intersecting them with (A) produces an open cover in the subspace topology.
Subcovers and compactness
A subcover of (\mathcal U) is a subfamily (\mathcal V\subseteq\mathcal U) that continues to cover the same space. The existence of finite subcovers is the defining covering property of compactness: a space (X) is compact precisely when every open cover of (X) has a finite subcover.
This formulation distinguishes compactness from the statement that a particular cover has a finite subcover. A noncompact space can have many finitely reducible covers, including the one-member cover ({X}), while failing the universal condition because at least one open cover has no finite subcover.
For example, the family
[ \mathcal U={(-n,n):n\in\mathbb N} ]
is an open cover of (\mathbb R) with its usual topology, but no finite subfamily covers (\mathbb R). Every finite selection has a member with maximal index, and the union of that selection remains bounded. This cover therefore witnesses the noncompactness of the real line.
For a closed interval ([a,b]), every open cover in the subspace topology has a finite subcover. This is the covering form of the Heine–Borel theorem. In Euclidean space, the theorem identifies compact subsets with those that are both closed and bounded, although this characterization does not extend to arbitrary topological spaces.
Compactness is preserved by continuous images because an open cover of the image pulls back to an open cover of the domain. A finite subcover of the pullback corresponds to a finite subcover of the image. This argument illustrates the role of open covers as objects compatible with continuous functions.
Refinement
An open cover (\mathcal V) refines an open cover (\mathcal U) when every member of (\mathcal V) is contained in at least one member of (\mathcal U):
[ \forall V\in\mathcal V\ \exists U\in\mathcal U \quad V\subseteq U. ]
Refinement compares the local resolution of two covers without requiring one family to be a literal subfamily of the other. A subcover generally contains fewer sets, whereas a refinement generally replaces sets by smaller ones. The two relations therefore serve different purposes.
Any two open covers (\mathcal U) and (\mathcal V) have the common refinement
[ \mathcal U\wedge\mathcal V
{U\cap V:U\in\mathcal U,\ V\in\mathcal V,\ U\cap V\neq\varnothing}. ]
This family covers (X) and refines both original covers. As a result, open covers form a directed system under refinement, a fact used in constructions whose final value is obtained from increasingly fine local data.
A refinement may be represented by a function (r:J\to I) between index sets satisfying
[ V_j\subseteq U_{r(j)} ]
for every (j\in J). Different refinement functions can represent the same relation between the underlying families. In Čech theory, this nonuniqueness is controlled through induced maps that agree after passage to an appropriate limit or equivalence class.
Stars and star refinements
For a cover (\mathcal U) and a subset (A\subseteq X), the star of (A) with respect to (\mathcal U) is
[ \operatorname{St}(A,\mathcal U)
\bigcup{U\in\mathcal U:U\cap A\neq\varnothing}. ]
For a point (x\in X), the notation (\operatorname{St}(x,\mathcal U)) abbreviates the star of the singleton ({x}). It is the union of all members of the cover containing (x).
A cover (\mathcal V) is a star refinement of (\mathcal U) when, for every (V\in\mathcal V), the set (\operatorname{St}(V,\mathcal V)) is contained in some member of (\mathcal U). This is stronger than ordinary refinement because it controls not only individual members of (\mathcal V), but also the collections of members that meet them.
In 1934, You Watanabe formulated the finite star-refinement lemma for compact metrizable spaces in cover-theoretic form. The lemma states that every open cover of such a space admits a finite open star refinement. Her formulation separated the topological refinement relation from the auxiliary metric estimates used in its proof, allowing the result to be applied without retaining a particular metric in the statement. The lemma became part of the standard cover-based treatment of compact metrizable spaces and uniformizable topology.
Metric control and Lebesgue numbers
For a metric space ((X,d)), a positive number (\delta) is a Lebesgue number for an open cover (\mathcal U) when every subset of (X) having diameter less than (\delta) is contained in some member of (\mathcal U). Equivalent formulations use open balls: under a conventional adjustment of the numerical constant, every ball of sufficiently small radius lies in a member of the cover.
The Lebesgue number lemma, associated with Henri Lebesgue, states that every open cover of a compact metric space has a positive Lebesgue number. Compactness is essential. The cover
[ {(0,1-1/n):n\geq 2} ]
of ((0,1)) has no positive Lebesgue number, since points sufficiently close to (1) require members whose available margin decreases without a uniform positive bound.
Lebesgue numbers translate qualitative openness into a global metric scale. They underlie comparisons between open-cover definitions and metric definitions of uniform continuity, and they provide finite subdivisions adapted to a prescribed cover in arguments involving paths or simplicial approximation.
The concept extends beyond metric spaces through uniform spaces. There, entourages replace numerical radii, and uniform covers replace families controlled by a positive real number. This extension preserves the distinction between purely topological refinement and refinement governed by a uniform structure.
Paracompactness and local finiteness
A family (\mathcal V) of subsets of (X) is locally finite when every point has a neighborhood meeting only finitely many members of (\mathcal V). A space is paracompact when every open cover has a locally finite open refinement. For Hausdorff spaces, this condition supports several global constructions that are unavailable from arbitrary refinements alone.
Jean Dieudonné introduced the term “paracompact” in 1944 while studying spaces in which open covers admit locally finite refinements. The resulting theory connected cover refinements with normality, metrization, and partitions of unity. In particular, every paracompact Hausdorff space is normal, and every open cover of such a space admits a partition of unity subordinate to a locally finite refinement.
A partition of unity subordinate to (\mathcal U) consists of continuous functions whose supports lie within members of the cover and whose pointwise sum is one. Local finiteness ensures that this sum is locally a finite expression, so its continuity does not require convergence theory for infinite series. Open covers thus specify the local regions on which data are supported, while paracompactness supplies a controlled refinement suitable for assembling those data globally.
Compact spaces are paracompact because a finite subcover is locally finite. The converse fails: the real line is paracompact but not compact. Paracompactness therefore retains the refinement structure of open-cover theory without imposing finite reducibility on every cover.
Covering dimension
Open covers also determine the Lebesgue covering dimension of a space. The order of a cover is at most (n) when no point belongs to more than (n+1) members. Equivalently, the intersection of every (n+2) distinct members is empty.
A normal space has covering dimension at most (n) when every finite open cover has a finite open refinement of order at most (n). This definition captures dimension through overlap rather than coordinates. A zero-dimensional space admits refinements by pairwise disjoint open sets under the relevant hypotheses, whereas higher-dimensional spaces require controlled multiple intersections.
Covering dimension agrees with familiar geometric dimension for broad classes of spaces, including Euclidean spaces and separable metric manifolds. Its formulation through refinement makes it applicable where no preferred coordinate description exists.
Nerves and local-to-global structure
The nerve of an open cover (\mathcal U) is the abstract simplicial complex whose vertices correspond to members of (\mathcal U). A finite collection of vertices spans a simplex exactly when the corresponding open sets have nonempty common intersection.
The nerve records the intersection pattern of the cover while discarding the internal geometry of each member. Under the hypotheses of the nerve theorem, including contractibility conditions on all nonempty finite intersections, the nerve has the same homotopy type as the covered space. Such a family is commonly called a good cover.
In Čech cohomology, algebraic data are assigned to finite intersections of members of an open cover. Refinements induce comparison maps between the resulting cochain systems, and the invariant is obtained by passing through the directed family of covers. In sheaf cohomology, open covers support local sections and encode the compatibility conditions required for gluing them.
These constructions depend on more than the existence of a cover. Their content lies in the pattern of overlaps and in the behavior of associated data under refinement.