An accumulation point of a set is a point that can be approached arbitrarily closely by points of that set other than the point itself. Formally, a point of a topological space is an accumulation point of if every neighborhood of intersects . The point need not belong to . Accumulation points are fundamental in mathematical analysis and topology, where they describe the local distribution of sets and help characterize closure, convergence, and compactness.
Definition
Let be a subset of a topological space . A point is an accumulation point of if
for every neighborhood of . Equivalently, the condition may be checked only for open neighborhoods.
The removal of is essential. Without it, every member of would automatically satisfy the condition, regardless of whether any other points of occur nearby.
In a metric space , the definition becomes
Thus every open ball centered at must contain a point of distinct from . On the real line with its usual metric, this says that every interval contains such a point.
“Limit point” is a common synonym. “Cluster point” is also used, but terminology varies, particularly when discussing sequences rather than sets.
Examples
A set with one accumulation point
Consider
Its only accumulation point is . Every interval around contains for sufficiently large , even though .
Each point , by contrast, has a sufficiently small neighborhood containing no other member of . Adding to the set does not change its accumulation points.
An interval
For , the set of accumulation points is . Both endpoints are accumulation points despite lying outside , while every interior point is approached by other points of the interval.
Rational numbers
Every real number is an accumulation point of the set of rational numbers . Indeed, every nonempty open interval contains infinitely many rational numbers. The same statement holds for the irrational numbers.
Integers and finite sets
The integers have no accumulation points in : sufficiently small intervals around any real number contain at most one integer, and an interval avoiding that integer can be chosen when necessary.
Likewise, a finite subset of a metric space has no accumulation points. This statement requires care in more general topological spaces, where finite sets can have accumulation points.
Closure, isolated points, and boundary
A point is an adherent point of if every neighborhood intersects , without excluding the point itself. The set of all adherent points is the closure .
Writing for the set of accumulation points of , one has, in every topological space,
Consequently, is a closed set if and only if it contains all its accumulation points.
A point is an isolated point of if some neighborhood satisfies
Thus each point of is either isolated in or an accumulation point of , but not both. An accumulation point outside belongs to the closure without being a member of the set.
Accumulation points should also be distinguished from boundary points. A boundary point has every neighborhood intersecting both and . An isolated point of can be a boundary point without being an accumulation point; an interior point of an interval is an accumulation point without being a boundary point.
The ambient space matters. The set has as an accumulation point in , but has no accumulation points when regarded as a space in its own right with the subspace topology.
Characterization by sequences
In metric spaces, accumulation points can be characterized through limits of sequences:
For the forward direction, progressively smaller balls around provide the sequence. The points can be chosen distinct because every neighborhood of an accumulation point in a metric space contains infinitely many points of . Conversely, a sequence of points different from converging to meets every neighborhood of .
The same characterization holds in any first-countable space. First countability supplies a countable local neighborhood base; the condition, meaning that every singleton is closed, allows previously selected points to be excluded.
Accumulation of a set versus clustering of a sequence
A cluster point of a sequence is usually defined as a point whose every neighborhood contains terms at infinitely many indices. In a metric space, this is equivalent to being the limit of a subsequence.
This differs from an accumulation point of the sequence’s range. A constant sequence converges to and has as a cluster point, but its range is the singleton , which has no accumulation points in a metric space. Sequence clustering counts repeated occurrences; accumulation of a set concerns distinct locations.
Beyond first-countable spaces
In arbitrary topological spaces, sequences may not detect every accumulation point. The general replacement is a net:
A convergent net of this kind can be constructed by directing the neighborhoods of by reverse inclusion and choosing a point of in each neighborhood.
Separation assumptions and infinite neighborhoods
In metric spaces, “every neighborhood contains another point of ” is equivalent to “every neighborhood contains infinitely many points of .” More generally, this equivalence holds in spaces, including every Hausdorff space.
To see why, suppose a neighborhood contains only finitely many points of . Since finite sets are closed in a space, those points can be removed to obtain a smaller neighborhood containing no point of , contradicting accumulation.
Without the assumption, the two conditions can differ. For example, let have only and as open sets. Then is an accumulation point of , because its only open neighborhood contains . No neighborhood can contain infinitely many points of this finite set.
Some authors therefore distinguish ordinary accumulation points from -accumulation points, whose every neighborhood contains infinitely many points of the set.
Derived sets and perfect sets
The set of all accumulation points of is called the derived set. In a space it is closed, though this need not hold in an arbitrary topological space.
Derived sets can be iterated:
For a closed set in a space, each step removes the isolated points of the preceding set. The construction extends to ordinal numbers by taking intersections at limit stages and forms the basis of Cantor–Bendixson analysis.
A perfect set is a closed set with no isolated points; equivalently,
The Cantor set is a standard example. Every one of its points is approached by other points of the set, even though the set contains no nonempty open interval. In the real line, every nonempty perfect set is uncountable.
Compactness and the Bolzano–Weierstrass theorem
Accumulation points are closely related to compactness. Every infinite subset of a compact space has an accumulation point in that space.
A direct proof uses the open-cover definition. If an infinite set had no accumulation point, each point of the space would have an open neighborhood meeting in at most that point. A finite subcover would then force to be finite.
For metric spaces, the following are equivalent:
- The space is compact.
- Every infinite subset has an accumulation point in the space.
- Every sequence has a convergent subsequence whose limit belongs to the space.
Outside metric spaces, these conditions require separate treatment and are not all equivalent.
The Bolzano–Weierstrass theorem states that every bounded infinite subset of Euclidean space has an accumulation point in . Equivalently, every bounded sequence in has a convergent subsequence. This follows from the compactness of closed bounded subsets expressed by the Heine–Borel theorem.
Roles in analysis
Accumulation points specify where nontrivial limits can be studied. In the usual definition of a function limit as , the point is required to be an accumulation point of the function’s domain. Otherwise, there may be no domain points satisfying , making the limit condition vacuous rather than determining a unique value.
In complex analysis, accumulation points appear in the identity theorem. If two analytic functions on a connected open domain agree on a set with an accumulation point inside that domain, they agree throughout the domain. The location of the accumulation point is essential: accumulation only at a boundary point does not generally force this conclusion.