Closed sets, closure and limit points
On the line, a point is close to a set when the set has points within every distance of it. A topology has no distances, only open sets. Which points can no open set hold apart from a given set?
Where you are 6 in Topological spaces
Why this exists
In Open, closed and compact sets in R a set of real numbers is closed when it contains its limit points, and Theorem 4 there shows that this happens exactly when its complement is open; Metric spaces, a first look takes the complement condition as the definition. A topology has no distances, so “arbitrarily close” has to be said with open sets: a point is close to A when every open set around it meets A.
With that change, closed sets, closure, interior, boundary and limit points make sense in every topological space, and the line’s theorem becomes Theorem 4 below, though not every fact about the line survives (Examples 1 and 2).
Definition
Throughout, (X, \mathcal{T}) is a topological space and A \subseteq X. A neighborhood of x is an open set containing x, as in Topologies and open sets.
Closed set. A set C \subseteq X is closed if X \setminus C is open. A set both open and closed is clopen.
Interior. The interior \operatorname{int} A is the union of all open sets contained in A. It is open by (T2), it is the largest open set inside A, and x \in \operatorname{int} A exactly when some neighborhood of x lies inside A.
Closure. The closure \overline{A} is the intersection of all closed sets containing A, a nonempty family since X is one of them.
Limit point. A point x \in X is a limit point of A if every neighborhood of x contains a point of A other than x. It need not lie in A. A point x \in A that is not a limit point of A is an isolated point of A: some neighborhood U of x has U \cap A = \{x\}.
Boundary. The boundary of A is \partial A = \overline{A} \setminus \operatorname{int} A.
Dense set. A is dense in X if \overline{A} = X.
Intuition
The closure asks whether some open set can hold a point while missing the set. If none can, the point is in the closure. A limit point passes a stricter test: no open set holds it apart from the rest of A, with the point itself taken out of A.
In Figure 1 the dashed inner circle is not in A, yet every disk about one of its points reaches into A: these are limit points. The center is in A, but the disk of radius 1 about it meets A only there, so it is in the closure without being a limit point.
Results
Theorem 1 (closed sets). (a) \varnothing and X are closed. (b) The intersection of any nonempty family of closed sets is closed. (c) The union of finitely many closed sets is closed. (d) If \mathcal{T}' is finer than \mathcal{T}, every \mathcal{T}-closed set is \mathcal{T}'-closed.
Theorem 2 (closure). (a) \overline{A} is closed, contains A, and lies inside every closed set that contains A. (b) A is closed if and only if A = \overline{A}. (c) If A \subseteq B then \overline{A} \subseteq \overline{B}. (d) \overline{\overline{A}} = \overline{A}.
Theorem 3 (the neighborhood test). A point x lies in \overline{A} if and only if every neighborhood of x meets A. If \mathcal{B} is a basis for \mathcal{T}, this holds if and only if every B \in \mathcal{B} with x \in B meets A. Consequently X \setminus \overline{A} = \operatorname{int}(X \setminus A).
Theorem 4 (closure and limit points). \overline{A} is the union of A and the set of limit points of A. Hence A is closed if and only if it contains all of its limit points.
Theorem 5 (limit points in a metric space). Let (X, d) be a metric space with its metric topology. A point x is a limit point of A if and only if every ball B(x, r) contains infinitely many points of A. In particular, a finite subset of X has no limit points and is closed.
Theorem 6 (dense sets). A is dense in X if and only if A meets every nonempty open set.
Why it is true
Theorem 1. Proof. (a) X \setminus \varnothing = X and X \setminus X = \varnothing are open by (T1). (b) For closed C_i, i \in I \ne \varnothing, De Morgan’s law (Topologies and open sets) makes the complement of \bigcap_{i \in I} C_i equal to \bigcup_{i \in I} (X \setminus C_i), open by (T2). (c) Likewise the complement of C_1 \cup \cdots \cup C_n is \bigcap_{i = 1}^{n} (X \setminus C_i), open by Theorem 1 of Topologies and open sets. (d) If X \setminus C \in \mathcal{T}, then X \setminus C \in \mathcal{T}'. \square
Theorem 2. Proof. (a) \overline{A} is closed by Theorem 1(b), contains A because every set intersected does, and each closed set containing A is one of the sets intersected. (b) If A is closed, it is one of those sets, so \overline{A} \subseteq A \subseteq \overline{A}. The converse is (a). (c) A closed set containing B contains A, hence \overline{A} by (a); so their intersection \overline{B} contains \overline{A}. (d) Apply (b) to \overline{A}. \square
Theorem 3. Proof. If a neighborhood U of x misses A, then X \setminus U is closed and contains A, so \overline{A} \subseteq X \setminus U by Theorem 2(a), and x \notin \overline{A}. Conversely, if x \notin \overline{A}, then X \setminus \overline{A} is a neighborhood of x missing A.
Basis elements are open, so if every neighborhood of x meets A, every basis element containing x does. Conversely, a neighborhood U of x contains some B \in \mathcal{B} with x \in B \subseteq U (Theorem 2 of Bases and subbases), and if B meets A, so does U.
Finally, x \in \operatorname{int}(X \setminus A) exactly when some neighborhood of x misses A, that is, by the first part, when x \notin \overline{A}. \square
Theorem 4. Proof. If x is a limit point of A, every neighborhood of x meets A \setminus \{x\} \subseteq A, so x \in \overline{A} by Theorem 3. Conversely, let x \in \overline{A} with x \notin A. Every neighborhood of x meets A (Theorem 3), at a point that cannot be x, so x is a limit point. With A \subseteq \overline{A} this proves the equality, and Theorem 2(b) gives the rest. \square
Theorem 5. Proof. Balls are open (Theorem 2(a) of Metric spaces, a first look), and every neighborhood of x contains a ball B(x, r). If every such ball holds infinitely many points of A, it holds one other than x, so x is a limit point.
Conversely, suppose B(x, r) holds only finitely many points of A. If none differs from x, this ball shows that x is not a limit point. Otherwise let a_1, \ldots, a_k be those other than x and s = \min\{d(x, a_1), \ldots, d(x, a_k)\} > 0; the neighborhood B(x, s) holds no point of A other than x. A finite set meets every ball in finitely many points, so it has no limit points, and it is closed by Theorem 4. \square
Theorem 6. Proof. Let \overline{A} = X and let U be open with a point x. Then U is a neighborhood of x \in \overline{A}, so it meets A (Theorem 3). Conversely, if A meets every nonempty open set, every neighborhood of every x meets A, and x \in \overline{A} by Theorem 3. \square
Worked examples
Cameras in an aviary
Amara runs an aviary with five enclosures a, b, c, d, e. Four cameras watch them: S_1 sees \{a, b, c\}, S_2 sees \{b\}, S_3 sees \{c, d\} and S_4 sees \{d, e\} (Figure 2). Take the four views as a subbasis on the set G of enclosures. (a) Find the smallest open set M_x containing each enclosure, and count the open sets. (b) For A = \{b, d\}, find \overline{A}, the limit points of A, \operatorname{int} A and \partial A. (c) Find the smallest dense subset of G.
(a) The views meet in S_1 \cap S_2 = \{b\}, S_1 \cap S_3 = \{c\}, S_3 \cap S_4 = \{d\}, and otherwise in \varnothing. As in Example 4 of Bases and subbases, M_x is the intersection of the views containing x:
| Enclosure | Cameras seeing it | M_x |
|---|---|---|
| a | S_1 | \{a, b, c\} |
| b | S_1, S_2 | \{b\} |
| c | S_1, S_3 | \{c\} |
| d | S_3, S_4 | \{d\} |
| e | S_4 | \{d, e\} |
A set U is open exactly when M_x \subseteq U for every x \in U (Example 6 of Topologies and open sets). So U \cap \{b, c, d\} is any of the eight subsets of \{b, c, d\}, a may join when b, c \in U, and e when d \in U. Over the eight choices, in the order \varnothing, \{b\}, \{c\}, \{d\}, \{b, c\}, \{b, d\}, \{c, d\}, \{b, c, d\}, this gives 1 + 1 + 1 + 2 + 2 + 2 + 2 + 4 = 15 open sets.
(b) Every neighborhood of x contains M_x, and M_x is one. So by Theorem 3, x \in \overline{A} exactly when M_x meets A, and x is a limit point exactly when M_x meets A \setminus \{x\}. M_a and M_e meet A at b and d; M_b = \{b\} and M_d = \{d\} meet A only at themselves; M_c = \{c\} misses A. Hence
the limit points of A are a and e, and b, d are isolated, as Theorem 4 requires. Since M_b, M_d \subseteq A, \operatorname{int} A = A: the set is open but not closed, and \partial A = \{a, e\}. Only c can be watched, by S_1 and S_3 together, without watching b or d. Theorem 5 fails here: M_a holds one point of A.
(c) Every nonempty open set contains some M_x, so by Theorem 6 a set is dense exactly when it meets every M_x. It must contain b, c and d, since M_b, M_c, M_d are single points, and \{b, c, d\} meets M_a and M_e too. The smallest dense set is \{b, c, d\}.
Closed in which topology?
On \mathbb{R} consider the standard topology, the lower limit topology \mathbb{R}_{\ell} and the finite-complement topology \mathcal{T}_f. In each, decide whether these sets are open, closed, both or neither: A = [-1, 4), B = \mathbb{Z}, C = \{0, 7, 11\} and D = (-\infty, 2].
SolutionThe closed sets of \mathcal{T}_f are \mathbb{R} and the finite sets. The standard topology is finer than \mathcal{T}_f (Theorem 4(b) of Topologies and open sets), and \mathbb{R}_{\ell} is finer still (Theorem 5(b) of Bases and subbases). So by Theorem 1(d) a set closed in one is closed in each finer one, and a set not open in \mathbb{R}_{\ell} is open in neither coarser topology.
C is finite, so it is closed in all three, and open in none, as [0, \varepsilon) \not\subseteq C.
B and D have the standard-open complements \bigcup_{n \in \mathbb{Z}} (n, n + 1) and (2, \infty), so they are closed in the standard topology and in \mathbb{R}_{\ell}, but not in \mathcal{T}_f, being infinite. Neither is open even in \mathbb{R}_{\ell}: every [0, \varepsilon) contains a non-integer, and every [2, 2 + \varepsilon) contains \displaystyle 2 + \frac{\varepsilon}{2} \notin D.
In the standard topology A is not closed: 4 \notin A, while every (4 - r, 4 + r) contains \displaystyle 4 - \frac{1}{2}\min\{r, 5\} \in A, so 4 \in \overline{A} (Theorem 3). It is not open: every (-1 - r, -1 + r) contains \displaystyle -1 - \frac{r}{2} \notin A. In \mathbb{R}_{\ell}, A is a basis element, and its complement (-\infty, -1) \cup [4, \infty) is open as well, with [4, \infty) = \bigcup_{n \ge 1} [4, 4 + n). So A is clopen in \mathbb{R}_{\ell}. In \mathcal{T}_f it is neither, as A and its complement are infinite.
| Set | standard | \mathbb{R}_{\ell} | \mathcal{T}_f |
|---|---|---|---|
| A | neither | clopen | neither |
| B | closed | closed | neither |
| C | closed | closed | closed |
| D | closed | closed | neither |
An annulus with its center
In \mathbb{R}^2 let A = \{x \in \mathbb{R}^2 : 1 < \lVert x \rVert \le 2\} \cup \{(0, 0)\} (Figure 1). Find the limit points and isolated points of A, and \overline{A}, \operatorname{int} A and \partial A.
SolutionLet K = \{x : 1 \le \lVert x \rVert \le 2\}. For p \in K, r > 0 and u = p / \lVert p \rVert, each su with 1 < s \le 2 lies in A at distance |s - \lVert p \rVert| from p. The s \in (1, 2] with |s - \lVert p \rVert| < r fill an interval of positive length, so one of them has s \ne \lVert p \rVert, and su \in B(p, r). Every point of K is a limit point.
No other point is. If 0 < \lVert p \rVert < 1 and s = \min\{\lVert p \rVert, 1 - \lVert p \rVert\}, the triangle inequality gives 0 < \lVert q \rVert < 1 for every q \in B(p, s), so this ball misses A; if \lVert p \rVert > 2, so does B(p, \lVert p \rVert - 2). And B\big((0, 0), 1\big) \cap A = \{(0, 0)\}, so the center is an isolated point, the only one. By Theorem 4, \overline{A} = K \cup \{(0, 0)\}.
If 1 < \lVert p \rVert < 2, the ball of radius \min\{\lVert p \rVert - 1, 2 - \lVert p \rVert\} about p lies in A. If \lVert p \rVert = 2, each B(p, r) contains \displaystyle \left(1 + \frac{r}{4}\right)p, of norm \displaystyle 2 + \frac{r}{2}, and each ball about the center contains \displaystyle \left(\frac{1}{2}\min\{r, 1\}, 0\right) \notin A. So \operatorname{int} A = \{x : 1 < \lVert x \rVert < 2\}, and \partial A is the two circles \lVert x \rVert = 1 and \lVert x \rVert = 2 together with (0, 0).
One set, three topologies
Find the closures of P = (0, 1) and \displaystyle H = \left\{-\frac{1}{n} : n \ge 1\right\} in the standard topology, in \mathbb{R}_{\ell} and in the finite-complement topology on \mathbb{R}.
SolutionStandard. \overline{P} = [0, 1]: this set is closed, and 0, 1 are limit points of P. The complement of H \cup \{0\} is the union of the open intervals (-\infty, -1), (0, \infty) and \displaystyle \left(-\frac{1}{n}, -\frac{1}{n + 1}\right), and 0 is a limit point of H, as (-r, r) contains \displaystyle -\frac{1}{n} once \displaystyle n > \frac{1}{r}. So \overline{H} = H \cup \{0\}.
Lower limit. By Theorem 3 with the basis of intervals [a, b), x \in \overline{S} exactly when every [x, x + \varepsilon) meets S. A standard closure is closed in \mathbb{R}_{\ell} (Theorem 1(d)), so the closure in \mathbb{R}_{\ell} lies inside it. Every [0, \varepsilon) meets P and [1, 2) misses it, so \overline{P} = [0, 1). The interval [0, 1) misses H, so \overline{H} = H: the set is closed in \mathbb{R}_{\ell}, not in the standard line.
Finite complement. The only closed set containing an infinite set is \mathbb{R}, so \overline{P} = \overline{H} = \mathbb{R}.
A finer topology has more closed sets, so closures shrink.
The closure of a union
Prove that \overline{A \cup B} = \overline{A} \cup \overline{B} for any subsets A, B of any space. Then show that this fails for an infinite union.
SolutionProof. Since A \subseteq A \cup B, Theorem 2(c) gives \overline{A} \subseteq \overline{A \cup B}, and likewise for B. Conversely, \overline{A} \cup \overline{B} is closed by Theorem 1(c) and contains A \cup B, so it contains \overline{A \cup B} by Theorem 2(a). \square
By induction the same holds for finitely many sets. In \mathbb{R} every singleton \{q\} is closed (Theorem 5), so
while \overline{\mathbb{Q}} = \mathbb{R}, since every interval contains a rational (The Archimedean property and density of the rationals). A union of infinitely many closed sets need not be closed.
A set of sums of reciprocals
Let \displaystyle A = \left\{\frac{1}{m} + \frac{1}{n} : m, n \ge 1\right\} \subseteq \mathbb{R}. Find the limit points of A and its closure.
SolutionEvery point of A lies in (0, 2], and each \displaystyle \frac{1}{k} is in A, as \displaystyle \frac{1}{2k} + \frac{1}{2k}.
Limit points found. The points \displaystyle \frac{1}{k} + \frac{1}{n} of A differ from \displaystyle \frac{1}{k} and lie within \displaystyle \frac{1}{n} of it, so each \displaystyle \frac{1}{k} is a limit point. So is 0: the points \displaystyle \frac{2}{n} \in A lie in (-r, r) once \displaystyle n > \frac{2}{r}.
No others. If x < 0 or x > 2, the open set (-\infty, 0) or (2, \infty) contains x and misses A. Let 0 < x \le 2 with \displaystyle x \ne \frac{1}{k} for every k, and write \displaystyle S_k = \left\{\frac{1}{k} + \frac{1}{n} : n \ge 1\right\}. If \displaystyle \frac{1}{m} + \frac{1}{n} > \frac{x}{2}, the larger of \displaystyle \frac{1}{m}, \displaystyle \frac{1}{n} exceeds \displaystyle \frac{x}{4}, so \displaystyle \min\{m, n\} < \frac{4}{x}. Fix an integer \displaystyle N \ge \frac{4}{x}; as A is symmetric in m and n,
For k \le N let \displaystyle r_k = \frac{1}{2}\left|x - \frac{1}{k}\right| > 0. A point \displaystyle \frac{1}{k} + \frac{1}{n} with \displaystyle \frac{1}{n} < r_k is at distance at least \displaystyle 2r_k - \frac{1}{n} > r_k from x, so B(x, r_k) holds only points of S_k with \displaystyle n \le \frac{1}{r_k}, finitely many. Let \displaystyle r = \min\left\{\frac{x}{2}, r_1, \ldots, r_N\right\}. Then \displaystyle B(x, r) \subseteq \left(\frac{x}{2}, \infty\right), so its points of A lie in S_1 \cup \cdots \cup S_N, finitely many from each, and x is not a limit point (Theorem 5).
So the limit points of A are 0 and the numbers \displaystyle \frac{1}{k}, k \ge 1. All but 0 lie in A, so by Theorem 4
Boundaries
- Closed is not the opposite of open. A set can be both (Example 2 in \mathbb{R}_{\ell}, every set in a discrete space) or neither (\mathbb{Q} in \mathbb{R}).
- Infinite unions of closed sets need not be closed (Example 5).
- Theorem 5 needs more than a topology (Example 1). Exercise 12 restores it whenever every one-point set is closed.
- Closure depends on the topology (Example 4) and on the surrounding space (Subspace and product topologies).
Where this goes next
A map is continuous exactly when the preimage of every closed set is closed; Continuous maps and homeomorphisms builds on this. Hausdorff spaces and separation asks for disjoint neighborhoods of distinct points, which makes every point a closed set and limits of sequences unique. In Connectedness, a space is connected when \varnothing and X are its only clopen sets.
Exercises
Finite spaces
- On X = \{1, 2, 3, 4, 5\} let the open sets be \varnothing, X, \{2\}, \{4, 5\}, \{2, 4, 5\}, \{1, 2\} and \{1, 2, 4, 5\}. (a) List the closed sets. (b) For A = \{1, 5\}, find \overline{A}, the limit points of A, \operatorname{int} A and \partial A. (c) Is \{2, 4\} dense?
- Kofi adds a wing of four enclosures p, q, r, s, watched by three cameras that see \{p, q\}, \{q, r\} and \{r, s\}; take these as a subbasis. (a) Find the smallest open set containing each enclosure, and count the closed sets. (b) Find \overline{\{p\}}, \overline{\{q\}} and the limit points of \{q\}. (c) Find the smallest dense set.
- Let X have at least two points and \varnothing \ne A \subsetneq X. Find the closed sets, \overline{A} and the limit points of A (a) in the discrete topology; (b) in the indiscrete topology.
Topologies on the line and the plane
- Let \mathcal{T} be the topology on \mathbb{R} whose open sets are \varnothing, \mathbb{R} and the rays (-\infty, a), a \in \mathbb{R} (Exercise 3 of Topologies and open sets). (a) Find the closed sets. (b) Find the closure and the limit points of \{5\} and of (0, 1). (c) Which subsets are dense?
- In \mathbb{R}_{\ell}, find the closure of (a) (-\infty, 3); (b) (2, 5); (c) \displaystyle \left\{5 + \frac{1}{n} : n \ge 1\right\}; (d) \displaystyle \left\{5 - \frac{1}{n} : n \ge 1\right\}.
- In \mathbb{R}^2 let Q be the union of \{(x, y) : |x| < 1,\ |y| \le 1\} and the segment \{(x, 0) : 2 \le x \le 3\}. Find \overline{Q}, \operatorname{int} Q, \partial Q and the isolated points of Q.
- In the finite-complement topology on \mathbb{Z}, find the closure and the limit points of \{-2, 0, 9\} and of the even integers. Which subsets of \mathbb{Z} are dense?
Proofs
- Prove that \overline{A \cap B} \subseteq \overline{A} \cap \overline{B}, and find intervals A, B in \mathbb{R} for which the inclusion is strict.
- Prove that \partial A = \overline{A} \cap \overline{X \setminus A}, that \partial A is closed, and that \partial A = \varnothing if and only if A is clopen.
- Let U be open and D dense in X. Prove that U \subseteq \overline{U \cap D}.
- Let (X, d) be a metric space. Prove that x \in \overline{A} if and only if some sequence of points a_n \in A converges to x (convergence as in Metric spaces, a first look).
- Suppose every one-point set \{y\} is closed in X. Prove that if x is a limit point of A, every neighborhood of x contains infinitely many points of A.
- Prove that \mathbb{Q} and \mathbb{R} \setminus \mathbb{Q} are both dense in \mathbb{R}_{\ell}, and deduce that they are dense in \mathbb{R}.
- Refute with a counterexample: “If \mathcal{S} is a subbasis, then x \in \overline{A} whenever every S \in \mathcal{S} with x \in S meets A.”
Answers
- (a) X, \{1, 3, 4, 5\}, \{1, 2, 3\}, \{1, 3\}, \{3, 4, 5\}, \{3\}, \varnothing. (b) The smallest open sets are M_1 = \{1, 2\}, M_2 = \{2\}, M_3 = X, M_4 = M_5 = \{4, 5\}. So \overline{A} = \{1, 3, 4, 5\}; the limit points are 3 and 4; \operatorname{int} A = \varnothing and \partial A = \{1, 3, 4, 5\}. (c) Yes: \{2, 4\} meets every M_x (Theorem 6).
- (a) The views meet in \{q\} and \{r\}, so M_p = \{p, q\}, M_q = \{q\}, M_r = \{r\}, M_s = \{r, s\}. An open set holds \varnothing, \{q\} or \{p, q\} of p, q and \varnothing, \{r\} or \{r, s\} of r, s: 9 open sets, so 9 closed sets. (b) Only M_p contains p, so \overline{\{p\}} = \{p\}. M_p and M_q contain q, so \overline{\{q\}} = \{p, q\}, and the only limit point of \{q\} is p. (c) \{q, r\}: a dense set meets M_q and M_r, and \{q, r\} meets all four.
- (a) Every set is closed, \overline{A} = A, and there are no limit points, since \{x\} is a neighborhood of x. (b) The closed sets are \varnothing and X, so \overline{A} = X. The only neighborhood of x is X, so x is a limit point when A \setminus \{x\} \ne \varnothing: every point if A has two or more points, every point but a if A = \{a\}.
- (a) \varnothing, \mathbb{R} and the rays [a, \infty). (b) The closed sets containing 5 are \mathbb{R} and the [a, \infty) with a \le 5, so \overline{\{5\}} = [5, \infty); its limit points are the x > 5, as every (-\infty, a) with a > x contains 5. Likewise \overline{(0, 1)} = [0, \infty), and every x \ge 0 is a limit point: for a > x \ge 0, (-\infty, a) \cap (0, 1) is a nonempty open interval. (c) The sets meeting every ray (-\infty, a) (Theorem 6): the sets not bounded below.
- (a) (-\infty, 3), closed since [3, \infty) = \bigcup_{n \ge 1} [3, 3 + n) is open. (b) [2, 5): every [2, 2 + \varepsilon) meets the set, [5, 6) misses it, and the standard closure [2, 5] is closed in \mathbb{R}_{\ell}. (c) The set with 5 added: [5, 5 + \varepsilon) contains \displaystyle 5 + \frac{1}{n} for \displaystyle n > \frac{1}{\varepsilon}. (d) The set itself: its standard closure adds only 5, and [5, 6) misses it.
- \overline{Q} is the closed square [-1, 1] \times [-1, 1] with the segment; \operatorname{int} Q = (-1, 1) \times (-1, 1); \partial Q is the square’s perimeter with the segment; Q has no isolated points. The segment has empty interior, as every disk about (x, 0) contains \displaystyle \left(x, \frac{r}{2}\right).
- The closed sets are \mathbb{Z} and the finite sets. \{-2, 0, 9\} is closed, its own closure, with no limit points: x has the neighborhood \mathbb{Z} \setminus (\{-2, 0, 9\} \setminus \{x\}). The even integers have closure \mathbb{Z}, and every integer is a limit point, as a neighborhood omits only finitely many integers. The dense sets are exactly the infinite subsets.
- A \cap B \subseteq A gives \overline{A \cap B} \subseteq \overline{A} by Theorem 2(c), and likewise for B. For A = (0, 1), B = (1, 2): \overline{A \cap B} = \varnothing, while \overline{A} \cap \overline{B} = [0, 1] \cap [1, 2] = \{1\}. \square
- Theorem 3 applied to X \setminus A gives X \setminus \overline{X \setminus A} = \operatorname{int} A, so \partial A = \overline{A} \cap (X \setminus \operatorname{int} A) = \overline{A} \cap \overline{X \setminus A}, closed by Theorem 1(b). If \partial A = \varnothing, then \overline{A} \subseteq \operatorname{int} A \subseteq A \subseteq \overline{A}, so A = \operatorname{int} A is open and A = \overline{A} is closed. If A is clopen, \overline{A} = A = \operatorname{int} A and \partial A = \varnothing. \square
- Let x \in U and V be a neighborhood of x. Then U \cap V is open and contains x, so it meets D (Theorem 6); that is, V meets U \cap D. By Theorem 3, x \in \overline{U \cap D}. \square
- If x \in \overline{A}, each ball \displaystyle B\left(x, \frac{1}{n}\right), n \ge 1, holds some a_n \in A (Theorem 3). Given \varepsilon > 0, take \displaystyle N > \frac{1}{\varepsilon}; then \displaystyle d(a_n, x) < \frac{1}{n} < \varepsilon for n \ge N. Conversely, if a_n \to x and U is a neighborhood of x, some B(x, \varepsilon) \subseteq U holds a_n for large n, so U meets A and x \in \overline{A}. \square
- Suppose a neighborhood U of x held only finitely many points of A, and let F be those other than x. F is closed by Theorem 1(a) and (c), so U \cap (X \setminus F) is a neighborhood of x holding no point of A other than x, and x is not a limit point. \square
- A nonempty open set of \mathbb{R}_{\ell} contains some [a, b) with a < b. By density of the rationals there is a rational q \in (a, b), and for \displaystyle n > \frac{\sqrt{2}}{b - q} the irrational number \displaystyle q + \frac{\sqrt{2}}{n} lies in (q, b). Theorem 6 gives both densities. The standard closure is \mathbb{R}_{\ell}-closed (Theorem 1(d)), so it contains the \mathbb{R}_{\ell} closure, \mathbb{R}. \square
- On X = \{1, 2, 3\} take \mathcal{S} = \{\{1, 2\}, \{2, 3\}\}, A = \{1, 3\} and x = 2. Both members of \mathcal{S} containing 2 meet A, but the basis element \{1, 2\} \cap \{2, 3\} = \{2\} is a neighborhood of 2 missing A, so 2 \notin \overline{A}. Theorem 3 needs a basis.