For a subset of a Euclidean space, is a point of closure of if every open ball centered at contains a point of (this point may be itself).
This definition generalizes to any subset of a metric space
Fully expressed, for a metric space with metric is a point of closure of if for every there exists some such that the distance (again, is allowed).
Another way to express this is to say that is a point of closure of if the distance
This definition generalizes to topological spaces by replacing "open ball" or "ball" with "neighbourhood".
Let be a subset of a topological space
Then is a point of closure or adherent point of if every neighbourhood of contains a point of 
Note that this definition does not depend upon whether neighbourhoods are required to be open.
The definition of a point of closure is closely related to the definition of a limit point of a set.
The difference between the two definitions is subtle but important – namely, in the definition of limit point, every neighbourhood of the point in question must contain a point of the set other than itself.
The set of all limit points of a set is called the derived set of
Thus, every limit point is a point of closure, but not every point of closure is a limit point.
A point of closure which is not a limit point is an isolated point.
In other words, a point is an isolated point of if it is an element of and if there is a neighbourhood of which contains no other points of other than itself.
For a given set and point is a point of closure of if and only if is an element of or is a limit point of (or both).
The closure of a subset of a topological space denoted by or possibly by (if is understood), where if both and are clear from context then it may also be denoted by or (moreover, is sometimes capitalized to ) can be defined using any of the following equivalent definitions:
Note that these properties are also satisfied if "closure", "superset", "intersection", "contains/containing", "smallest" and "closed" are replaced by "interior", "subset", "union", "contained in", "largest", and "open".
For more on this matter, see closure operator below.
Consider a sphere in 3 dimensions. Implicitly there are two regions of interest created by this sphere; the sphere itself and its interior (which is called an open 3-ball). It is useful to be able to distinguish between the interior of 3-ball and the surface, so we distinguish between the open 3-ball, and the closed 3-ball – the closure of the 3-ball. The closure of the open 3-ball is the open 3-ball plus the surface.
If one considers on the trivial topology in which the only closed (open) sets are the empty set and itself, then
These examples show that the closure of a set depends upon the topology of the underlying space. The last two examples are special cases of the following.
In any discrete space, since every set is closed (and also open), every set is equal to its closure.
In any indiscrete space since the only closed sets are the empty set and itself, we have that the closure of the empty set is the empty set, and for every non-empty subset of In other words, every non-empty subset of an indiscrete space is dense.
The closure of a set also depends upon in which space we are taking the closure. For example, if is the set of rational numbers, with the usual relative topology induced by the Euclidean space and if then is both closed and open in because neither nor its complement can contain , which would be the lower bound of , but cannot be in because is irrational. So, has no well defined closure due to boundary elements not being in . However, if we instead define to be the set of real numbers and define the interval in the same way then the closure of that interval is well defined and would be the set of all real numbers greater than or equal to.
A closure operator on a set is a mapping of the power set of , into itself which satisfies the Kuratowski closure axioms.
Given a topological space, the topological closure induces a function that is defined by sending a subset to where the notation or may be used instead. Conversely, if is a closure operator on a set then a topological space is obtained by defining the closed sets as being exactly those subsets that satisfy (so complements in of these subsets form the open sets of the topology).
The closure operator is dual to the interior operator, which is denoted by in the sense that
Therefore, the abstract theory of closure operators and the Kuratowski closure axioms can be readily translated into the language of interior operators by replacing sets with their complements in
In general, the closure operator does not commute with intersections. However, in a complete metric space the following result does hold:
The closure of an intersection of sets is always a subset of (but need not be equal to) the intersection of the closures of the sets.
In a union of finitely many sets, the closure of the union and the union of the closures are equal; the union of zero sets is the empty set, and so this statement contains the earlier statement about the closure of the empty set as a special case.
The closure of the union of infinitely many sets need not equal the union of the closures, but it is always a superset of the union of the closures.
If and if is a subspace of (meaning that is endowed with the subspace topology that induces on it), then and the closure of computed in is equal to the intersection of and the closure of computed in :
Because is a closed subset of the intersection is a closed subset of (by definition of the subspace topology), which implies that (because is the smallest closed subset of containing ). Because is a closed subset of from the definition of the subspace topology, there must exist some set such that is closed in and Because and is closed in the minimality of implies that Intersecting both sides with shows that
It follows that is a dense subset of if and only if is a subset of
It is possible for to be a proper subset of for example, take and
If but is not necessarily a subset of then only
is always guaranteed, where this containment could be strict (consider for instance with the usual topology, and [proof 1]), although if happens to an open subset of then the equality will hold (no matter the relationship between and ).
Let and assume that is open in Let which is equal to (because ). The complement is open in where being open in now implies that is also open in Consequently is a closed subset of where contains as a subset (because if is in then ), which implies that Intersecting both sides with proves that The reverse inclusion follows from
Consequently, if is any open cover of and if is any subset then:
because for every (where every is endowed with the subspace topology induced on it by ).
This equality is particularly useful when is a manifold and the sets in the open cover are domains of coordinate charts.
In words, this result shows that the closure in of any subset can be computed "locally" in the sets of any open cover of and then unioned together.
In this way, this result can be viewed as the analogue of the well-known fact that a subset is closed in if and only if it is "locally closed in ", meaning that if is any open cover of then is closed in if and only if is closed in for every
A function between topological spaces is continuous if and only if the preimage of every closed subset of the codomain is closed in the domain; explicitly, this means: is closed in whenever is a closed subset of
In terms of the closure operator, is continuous if and only if for every subset
That is to say, given any element that belongs to the closure of a subset necessarily belongs to the closure of in If we declare that a point is close to a subset if then this terminology allows for a plain English description of continuity: is continuous if and only if for every subset maps points that are close to to points that are close to Thus continuous functions are exactly those functions that preserve (in the forward direction) the "closeness" relationship between points and sets: a function is continuous if and only if whenever a point is close to a set then the image of that point is close to the image of that set.
Similarly, is continuous at a fixed given point if and only if whenever is close to a subset then is close to
A function is a (strongly) closed map if and only if whenever is a closed subset of then is a closed subset of
In terms of the closure operator, is a (strongly) closed map if and only if for every subset
Equivalently, is a (strongly) closed map if and only if for every closed subset
One may elegantly define the closure operator in terms of universal arrows, as follows.
The powerset of a set may be realized as a partial ordercategory in which the objects are subsets and the morphisms are inclusion maps whenever is a subset of Furthermore, a topology on is a subcategory of with inclusion functor The set of closed subsets containing a fixed subset can be identified with the comma category This category — also a partial order — then has initial object Thus there is a universal arrow from to given by the inclusion
Similarly, since every closed set containing corresponds with an open set contained in we can interpret the category as the set of open subsets contained in with terminal object the interior of
All properties of the closure can be derived from this definition and a few properties of the above categories. Moreover, this definition makes precise the analogy between the topological closure and other types of closures (for example algebraic closure), since all are examples of universal arrows.