Derived Set

Basis
Last updated: Tags: Topology

Prerequisites

You already know what an accumulation point of a set is: a point that every open neighbourhood of reaches back into the set. Collecting all accumulation points of AA into a single set gives you the derived set, a compact package of the “limit behavior” of AA that turns out to be exactly the right tool for characterising closedness and building the closure.

Definition

Let (X,τ)(X, \tau) be a topological space and A⊆XA \subseteq X. The derived set of AA, written A′A', is the set of all accumulation points of AA:

A′≔{x∈X:∀ U∈τ, x∈U  ⟹  (U∖{x})∩A≠∅}.(1)A' \coloneqq \{ x \in X : \forall\, U \in \tau,\ x \in U \implies (U \setminus \{x\}) \cap A \neq \varnothing \}. \tag{1}

The derived set is a subset of XX, but it need not be a subset of AA, and AA need not be a subset of A′A'.

Examples

Standard topology on R\mathbb{R}

  • Let A=(0,1)A = (0, 1). Then A′=[0,1]A' = [0, 1]: every point in [0,1][0,1] is touched by (0,1)(0,1) in every open neighbourhood, including the endpoints 00 and 11 even though they are not in AA.
  • Let A={1/n:n∈N+}A = \{1/n : n \in \mathbb{N}^+\}. Then A′={0}A' = \{0\}: the only accumulation point is 00, and 0∉A0 \notin A.
  • Let A=QA = \mathbb{Q}. Then A′=RA' = \mathbb{R}: every real number is approachable by rationals.
  • Let A=ZA = \mathbb{Z}. Then A′=∅A' = \varnothing: every integer is isolated (it has a neighbourhood containing no other integer).

Discrete topology

In the discrete topology, every singleton is open, so no point can be an accumulation point of any set. Thus A′=∅A' = \varnothing for every A⊆XA \subseteq X.

Finite-complement topology on R\mathbb{R} (Zariski-like)

In the finite-complement topology on R\mathbb{R} — where a set is open iff its complement is finite — the open sets are large. For any infinite set AA and any x∈Rx \in \mathbb{R}, every open set U∋xU \ni x has finite complement, so UU contains all but finitely many points of R\mathbb{R}, and in particular (U∖{x})∩A(U \setminus \{x\}) \cap A is infinite. Thus A′=RA' = \mathbb{R} for every infinite AA.

Properties

Let (X,τ)(X, \tau) be a topological space and A,B⊆XA, B \subseteq X.

  • Monotone: if A⊆BA \subseteq B then A′⊆B′A' \subseteq B'. (More points in BB can only help fill open neighbourhoods.)
  • Union: (A∪B)′=A′∪B′(A \cup B)' = A' \cup B'. Every accumulation point of A∪BA \cup B is an accumulation point of AA or BB (or both).
  • Intersection: (A∩B)′⊆A′∩B′(A \cap B)' \subseteq A' \cap B', but equality need not hold.
  • Derived set of the derived set: A′′⊆A∪A′A'' \subseteq A \cup A', where A′′=(A′)′A'' = (A')' is the derived set of the derived set. This inclusion says that accumulation points of accumulation points of AA are themselves either in AA or accumulation points of AA.
  • A′A' is closed (proved below).

The derived set is closed

To see that A′A' is always a closed set, you need to show that its complement X∖A′X \setminus A' is open.

Take any point y∉A′y \notin A'. Since yy is not an accumulation point of AA, there exists an open set V∋yV \ni y with (V∖{y})∩A=∅(V \setminus \{y\}) \cap A = \varnothing — that is, V∩A⊆{y}V \cap A \subseteq \{y\}.

Now take any point z∈Vz \in V with z≠yz \neq y. You want to show z∉A′z \notin A', i.e., zz is not an accumulation point of AA. The open set VV is a neighbourhood of zz (since VV is open and z∈Vz \in V), and (V∖{z})∩A⊆(V∩A)∖{z}⊆{y}∖{z}=∅(V \setminus \{z\}) \cap A \subseteq (V \cap A) \setminus \{z\} \subseteq \{y\} \setminus \{z\} = \varnothing (since z≠yz \neq y). So z∉A′z \notin A'.

This shows every y∉A′y \notin A' has an open neighbourhood V∋yV \ni y contained in X∖A′X \setminus A', so X∖A′X \setminus A' is open and A′A' is closed.

Derived set and closedness

The derived set gives a clean characterisation of closed sets — arguably its most important application:

Theorem. A set A⊆XA \subseteq X is closed if and only if A′⊆AA' \subseteq A.

Proof sketch. (⇒\Rightarrow) Suppose AA is closed, so X∖AX \setminus A is open. If x∉Ax \notin A then X∖AX \setminus A is an open set containing xx with (X∖A)∩A=∅(X \setminus A) \cap A = \varnothing, so xx is not an accumulation point of AA. Hence A′⊆AA' \subseteq A.

(⇐\Leftarrow) Suppose A′⊆AA' \subseteq A. Take any y∉Ay \notin A; then y∉A′y \notin A', so there exists open U∋yU \ni y with U∩A⊆{y}U \cap A \subseteq \{y\}. Since y∉Ay \notin A, this gives U∩A=∅U \cap A = \varnothing, i.e., U⊆X∖AU \subseteq X \setminus A. So every point of X∖AX \setminus A has an open neighbourhood inside X∖AX \setminus A, making X∖AX \setminus A open and AA closed. □\square

In plain language: AA is closed exactly when it already “contains all its own limit behavior” — all the points that AA accumulates to are already in AA.

From derived set to closure

The derived set is the key ingredient in the formula for the closure of AA:

A‾=A∪A′.(2)\overline{A} = A \cup A'. \tag{2}

The closure A‾\overline{A} is the smallest closed set containing AA. Formula (2)(2) says you get it by adjoining to AA exactly those points outside AA that AA accumulates to. You can verify that A∪A′A \cup A' is closed using the theorem above: (A∪A′)′=A′∪A′′⊆A′∪(A∪A′)=A∪A′(A \cup A')' = A' \cup A'' \subseteq A' \cup (A \cup A') = A \cup A', confirming A∪A′A \cup A' contains its own derived set.

Iterating the derived set (Cantor–Bendixson)

Cantor first introduced the derived set precisely because iterating it reveals structure. Starting from AA, form A′=A(1)A' = A^{(1)}, then A′′=A(2)A'' = A^{(2)}, and so on. For subsets of R\mathbb{R}, this sequence eventually stabilises (possibly at ∅\varnothing) after countably many steps. The Cantor–Bendixson theorem uses this to decompose any closed subset of R\mathbb{R} into a perfect set and a countable set, a result with deep consequences in descriptive set theory.

Summary

  • The derived set A′A' is the collection of all accumulation points of AA.
  • A′A' need not be a subset of AA, and AA need not be a subset of A′A'.
  • A′A' is always a closed set.
  • AA is closed if and only if A′⊆AA' \subseteq A.
  • The closure satisfies A‾=A∪A′\overline{A} = A \cup A': you close up a set by adjoining its derived set.
  • The derived set is monotone and distributes over unions; iterating it reveals deep structural information (Cantor–Bendixson).