Topologies and open sets
In the line, the plane and every metric space, the open sets obey the same three rules, and continuity can be told from the open sets alone. What if those three rules, and nothing else, were all a space had?
Where you are 6 in Topological spaces
Why this exists
In a metric space (X, d) a set U is open when every point of U is the center of some ball B(x, r) \subseteq U. Theorem 2 of that lesson shows that \varnothing and X are open, that every union of open sets is open, and that every finite intersection of open sets is open. Continuity can be read off these sets alone: a function is continuous exactly when the preimage of every open set is open (Theorem 7 there). And the taxicab, Euclidean and maximum metrics on \mathbb{R}^n, whose balls have different shapes, have the same open sets, so the same continuous functions.
Topology keeps the open sets and drops the metric: a topology on a set is a chosen family of subsets, called open, obeying the three rules above and nothing more. Some useful topologies come from no metric at all (Theorem 5, Example 5), and very different metrics can give the same topology (Example 4).
Definition
Families of sets. Let I be any index set, finite or infinite, countable or not, and let U_i \subseteq X for each i \in I. The union \bigcup_{i \in I} U_i is the set of points in at least one U_i; for I \ne \varnothing the intersection \bigcap_{i \in I} U_i is the set of points in every U_i. These extend Union, intersection and complement, and De Morgan’s laws from Set identities: distributive and De Morgan laws hold for families with the same proof (x \notin \bigcup_{i \in I} U_i exactly when x \notin U_i for every i):
Topology. Let X be a set. A topology on X is a family \mathcal{T} \subseteq \mathcal{P}(X) of subsets of X (see Power set) such that:
- (T1) \varnothing \in \mathcal{T} and X \in \mathcal{T};
- (T2) for every index set I, if U_i \in \mathcal{T} for each i \in I, then \bigcup_{i \in I} U_i \in \mathcal{T};
- (T3) if U \in \mathcal{T} and V \in \mathcal{T}, then U \cap V \in \mathcal{T}.
The pair (X, \mathcal{T}) is a topological space, written X when the topology is clear. The members of \mathcal{T} are the open sets. An open set containing a point x is a neighborhood of x. Four topologies.
- The discrete topology: \mathcal{T} = \mathcal{P}(X).
- The indiscrete topology: \mathcal{T} = \{\varnothing, X\}.
- The finite-complement topology: \varnothing together with every U \subseteq X for which X \setminus U is finite.
- The metric topology of a metric d on X: U is open when every x \in U has some r > 0 with B(x, r) \subseteq U. The standard topology on \mathbb{R} is that of |x - y|, and on \mathbb{R}^n that of \lVert x - y \rVert.
A topology that is the metric topology of some metric on X is metrizable.
Comparing topologies. If \mathcal{T} \subseteq \mathcal{T}' are topologies on the same set, \mathcal{T}' is finer than \mathcal{T} and \mathcal{T} is coarser than \mathcal{T}', strictly if \mathcal{T} \ne \mathcal{T}'. Two topologies are comparable if one contains the other.
Intuition
A point of an open set in a metric space has room around it: move it a little and it stays in the set. A topology keeps the word “open” and lets the chosen family decide what “a little” means. A finer topology has more open sets and tells points apart more sharply. The discrete topology, where each \{x\} is open, separates every point from every other; the indiscrete topology separates none.
The rules are not symmetric: any union, but only finite intersections. In the plane the open disks \displaystyle B\big((0, 0), 1 + \frac{1}{n}\big), n \ge 1, intersect in the closed disk D = \{x \in \mathbb{R}^2 : \lVert x \rVert \le 1\} (Figure 1). A point with \lVert x \rVert \le 1 lies in all of them, and one with \lVert x \rVert > 1 misses those with \displaystyle \frac{1}{n} < \lVert x \rVert - 1. But D is not open: every ball about (1, 0) contains a point \displaystyle \big(1 + \frac{r}{2}, 0\big) outside D.
Finitely many positive radii have a positive least one; infinitely many can shrink to 0, as at (1, 0).
Results
Theorem 1 (finite intersections). In a topological space, the intersection U_1 \cap \cdots \cap U_n of finitely many open sets, n \ge 1, is open.
Theorem 2 (open is a local property). Let (X, \mathcal{T}) be a topological space and A \subseteq X. Then A is open if and only if every x \in A has a neighborhood U with U \subseteq A.
Theorem 3 (the named topologies). For every set X, the discrete, indiscrete and finite-complement families are topologies on X, and so is the metric topology of every metric on X. If X is finite, the finite-complement topology is the discrete topology.
Theorem 4 (comparison). (a) Every topology on X is finer than the indiscrete topology and coarser than the discrete topology. (b) On \mathbb{R}, the standard topology is strictly finer than the finite-complement topology.
Theorem 5 (metric topologies separate points). If \mathcal{T} is metrizable and x \ne y in X, then x and y have disjoint neighborhoods. Consequently the indiscrete topology on a set with at least two points, and the finite-complement topology on an infinite set, are not metrizable.
Theorem 6 (intersections of topologies). The intersection of any nonempty family of topologies on X is a topology on X. Hence for every family \mathcal{S} of subsets of X there is a coarsest topology containing \mathcal{S}: the intersection of all topologies on X that contain \mathcal{S}. A union of two topologies need not be a topology.
Why it is true
Theorem 1. Proof. By induction on n; the case n = 1 is trivial. If W = U_1 \cap \cdots \cap U_n is open, then U_1 \cap \cdots \cap U_{n + 1} = W \cap U_{n + 1} by the associative law, and this is open by (T3). \square
Theorem 2. Proof. If A is open, U = A serves for every x. Conversely, let W be the union of all open sets contained in A. Then W is open by (T2) and W \subseteq A; and each x \in A has a neighborhood inside A, one of the sets making up W, so A \subseteq W. Hence A = W is open. \square
Theorem 3. Proof. Discrete: unions and intersections of subsets of X are subsets of X. Indiscrete: a union of sets each equal to \varnothing or X is X if some member is X and \varnothing otherwise, and likewise for the intersection of two.
Finite complement. \varnothing is in by definition, and X \setminus X = \varnothing is finite. Let each U_i, i \in I, be open. If all are empty, so is the union. Otherwise some U_j \ne \varnothing has X \setminus U_j finite, and by De Morgan’s law
so the union has finite complement. For open U, V: if either is empty, so is U \cap V; otherwise X \setminus (U \cap V) = (X \setminus U) \cup (X \setminus V) is finite.
Metric: this is Theorem 2(b) of Metric spaces, a first look. If X is finite, every subset has finite complement, so every subset is open. \square
Theorem 4. Proof. (a) Every topology contains \varnothing and X by (T1), and all its members are subsets of X.
(b) Let U \ne \varnothing be open in the finite-complement topology, so F = \mathbb{R} \setminus U is finite. If F = \varnothing, U = \mathbb{R}. Otherwise, for x \in U let r = \min\{|x - a| : a \in F\}, positive as the least of finitely many positive numbers; then (x - r, x + r) misses F, so lies in U. Thus U is open in the standard topology. The inclusion is strict: (0, 1) is standard-open with infinite complement. \square
Theorem 5. Proof. Let d be a metric with topology \mathcal{T} and \displaystyle r = \frac{1}{2}d(x, y) > 0. The balls B(x, r) and B(y, r) are open (Theorem 2(a) of Metric spaces, a first look). A point z in both would give d(x, y) \le d(x, z) + d(z, y) < 2r = d(x, y), which is impossible.
In the indiscrete topology the only neighborhood of x is X, which contains y. In the finite-complement topology on an infinite X, nonempty open U, V have X \setminus (U \cap V) = (X \setminus U) \cup (X \setminus V) finite, so U \cap V is infinite, hence nonempty. \square
Theorem 6. Proof. Let \mathcal{T}_j, j \in J, be topologies on X, J \ne \varnothing, and \mathcal{T} = \bigcap_{j \in J} \mathcal{T}_j. Each \mathcal{T}_j contains \varnothing and X, so \mathcal{T} does. If every U_i lies in \mathcal{T}, then for each j all U_i lie in \mathcal{T}_j, so \bigcup_{i \in I} U_i \in \mathcal{T}_j by (T2); as this holds for every j, the union is in \mathcal{T}. The same argument with (T3) handles U \cap V.
The topologies containing \mathcal{S} form a nonempty family, as \mathcal{P}(X) is one. Their intersection is a topology, contains \mathcal{S}, and lies inside every topology containing \mathcal{S}. Example 2 gives two topologies whose union is not a topology. \square
Worked examples
Three families on four points
Let X = \{p, q, r, s\}. Which of these families are topologies on X?
In a finite family, any union of members is a union of finitely many distinct ones, built up two at a time, and the empty union is \varnothing; so it suffices to check unions and intersections of two members. A nested pair U \subseteq V gives V and U, so only pairs where neither contains the other need checking.
\mathcal{T}_A is a topology. The only such pair is \{p, q\}, \{p, r\}, with union \{p, q, r\} and intersection \{p\}, both present.
\mathcal{T}_B is not: \{q\} \cup \{r\} = \{q, r\} is missing, so (T2) fails.
\mathcal{T}_C is not: \{p, q\} \cap \{q, r\} = \{q\} is missing, so (T3) fails.
The coarsest topology containing two sets
On X = \{p, q, r, s\}, find the coarsest topology \mathcal{T} containing \{q, r\} and \{r, s\}. Then find \mathcal{T}_A \cap \mathcal{T}, with \mathcal{T}_A from Example 1, and decide whether \mathcal{T}_A \cup \mathcal{T} is a topology.
SolutionA topology containing the two sets contains \varnothing and X by (T1), their intersection \{r\} by (T3) and their union \{q, r, s\} by (T2):
This is already a topology, as the only pair in which neither contains the other is \{q, r\}, \{r, s\}. So \mathcal{T}, with six open sets, is the coarsest.
Only \varnothing and X lie in both, so \mathcal{T}_A \cap \mathcal{T} is the indiscrete topology. The union is not a topology: it contains \{p\} and \{r, s\} but not \{p, r, s\}.
Two topologies on the real line
Let \mathcal{T}_f be the finite-complement topology on \mathbb{R} and \mathcal{T}_s the standard topology. In which of them is each set open: A = \mathbb{R} \setminus \{-1, 2\}, B = \mathbb{R} \setminus \mathbb{Z}, C = (0, \infty), and \displaystyle D = \mathbb{R} \setminus \left\{\frac{1}{n} : n \ge 1\right\}?
SolutionA has a two-point complement, so A \in \mathcal{T}_f, and so A \in \mathcal{T}_s by Theorem 4(b). B, C, D have infinite complements, so none is in \mathcal{T}_f.
B \in \mathcal{T}_s: for x \notin \mathbb{Z}, \lfloor x \rfloor < x < \lfloor x \rfloor + 1, so r = \min\{x - \lfloor x \rfloor, \lfloor x \rfloor + 1 - x\} > 0 and (x - r, x + r) contains no integer. C \in \mathcal{T}_s: for x > 0, (x - x, x + x) \subseteq C.
D \notin \mathcal{T}_s: 0 \in D, and for every r > 0 an integer \displaystyle n > \frac{1}{r} puts \displaystyle \frac{1}{n} \notin D inside (-r, r).
Each \displaystyle \mathbb{R} \setminus \left\{\frac{1}{n}\right\} is open in both topologies, yet their intersection over n \ge 1 is D, open in neither: Theorem 1 does not extend to infinitely many sets.
A strange metric with the standard topology
On \mathbb{R} let d(x, y) = |x^3 - y^3|. (a) Show that d is a metric. (b) Find B_d(1, 7) and B_d(1, 1). (c) Prove that the metric topology of d is the standard topology.
Solution(a) d \ge 0 and d(x, y) = d(y, x) are clear. Cubing is strictly increasing, so x^3 = y^3 only when x = y. The triangle inequality is that of |\cdot| applied to x^3, y^3, z^3.
(b) B_d(1, 7) = \{y : -6 < y^3 < 8\}, and the cube root is strictly increasing, so \displaystyle B_d(1, 7) = (-\sqrt[3]{6}, 2). Likewise \displaystyle B_d(1, 1) = \{y : 0 < y^3 < 2\} = (0, \sqrt[3]{2}). The first is lopsided: it reaches about 2.82 left of 1 but 1 right.
(c) Proof. It is enough that each kind of ball about a point a contains a ball of the other kind about a: then a set holding a ball of one kind about each of its points holds one of the other kind too. Given \varepsilon > 0, let
If |y^3 - a^3| < r then (a - \varepsilon)^3 < y^3 < (a + \varepsilon)^3, so |y - a| < \varepsilon. Conversely, \displaystyle B_d(a, r) = \left(\sqrt[3]{a^3 - r}, \sqrt[3]{a^3 + r}\right) is an open interval containing a, so it contains (a - \delta, a + \delta) for \delta the smaller distance from a to its ends. \square
The metrics disagree about distance, as d(0, y) = |y|^3, yet have the same open sets.
A watershed
A watershed has five sites W = \{a, b, c, d, e\}. Water flows from a to c, from b to c, from c to e and from d to e (Figure 2). Water from x reaches y if y = x or y lies downstream of x along the arrows; call U \subseteq W sealed if water from a site of U reaches only sites of U. (a) Show that the sealed sets form a topology \mathcal{T}_W. (b) List its open sets. (c) Find the smallest open sets containing a and c. (d) Is \mathcal{T}_W metrizable?
(a) \varnothing and W are sealed. If x lies in a union of sealed sets, it lies in one of them, U_j, and every site reached from x is in U_j, so in the union. If x \in U \cap V with U, V sealed, every site reached from x is in both.
(b) Water from every site reaches e, so every nonempty sealed set contains e; one containing a or b contains c; and d is free. The choices within \{a, b, c\} are \varnothing, \{c\}, \{a, c\}, \{b, c\}, \{a, b, c\}, each with or without d, and with \varnothing itself:
11 open sets.
(c) A sealed set containing a contains a, c, e, and \{a, c, e\} is sealed, so it is the smallest; likewise \{c, e\} is the smallest containing c.
(d) No. Every neighborhood of c contains e, so c and e have no disjoint neighborhoods (Theorem 5).
The smallest neighborhood in a finite space
Let (X, \mathcal{T}) be a topological space with X finite, and for x \in X let M_x be the intersection of all neighborhoods of x. Prove that M_x is the smallest neighborhood of x, and that U \subseteq X is open if and only if M_x \subseteq U for every x \in U.
SolutionProof. The neighborhoods of x form a nonempty family, containing X, and a finite one, as \mathcal{P}(X) is finite. So M_x is open by Theorem 1. It contains x and lies inside every neighborhood of x.
If U is open and x \in U, then U is one of the sets intersected, so M_x \subseteq U. Conversely, if M_x \subseteq U for each x \in U, each such x has the neighborhood M_x inside U, and U is open by Theorem 2. \square
In Example 5, M_x is the set of sites reached from x: the topology is the flow, recorded as open sets. On an infinite set M_x need not be open: in the standard topology on \mathbb{R}, M_0 = \{0\}.
Boundaries
- Arbitrary intersections fail in general (Figure 1, Example 3), though on a finite set they hold (Exercise 14).
- Not every topology comes from a metric (Theorem 5); the property it uses is named in Hausdorff spaces and separation. And a metric carries more than its topology: \min\{|x - y|, 1\} has the usual intervals as its balls of radius at most 1, so gives \mathbb{R} its standard topology, yet makes it bounded.
- Open depends on the topology and on the set. Subspace and product topologies shows how a subset inherits open sets of its own.
- Open and closed are not opposites. In a discrete space every set is both; Closed sets, closure and limit points takes this up.
Where this goes next
Bases and subbases describes a topology by a smaller family, as open intervals do for \mathbb{R}, and the coarsest topology containing a family (Theorem 6) becomes the topology generated by a subbasis. Closed sets, closure and limit points builds closure from open sets alone, Continuous maps and homeomorphisms takes the preimage criterion as the definition of continuity, and Compactness and open covers works with covers by open sets.
Exercises
Is it a topology?
- On X = \{1, 2, 3, 4, 5\}, which of these families are topologies? (a) \{\varnothing, \{2\}, \{2, 4\}, \{1, 2, 4\}, X\}; (b) \{\varnothing, \{1, 2\}, \{3, 4\}, \{1, 2, 3, 4\}, X\}; (c) \{\varnothing, \{1\}, \{2\}, \{1, 2, 3\}, X\}; (d) \{\{1\}, \{1, 3\}, X\}.
- On X = \{1, 2, 3, 4, 5\}, find the coarsest topology containing \{1, 2\}, \{2, 3, 4\} and \{4, 5\}. How many open sets does it have?
- Which of these families of subsets of \mathbb{R} are topologies? (a) \varnothing, \mathbb{R} and every ray (-\infty, a) with a \in \mathbb{R}; (b) \varnothing, \mathbb{R} and every ray (-\infty, a] with a \in \mathbb{R}; (c) \varnothing, \mathbb{R} and every ray (-\infty, a) with a \in \mathbb{Q}.
Open sets in particular spaces
- In the finite-complement topology on \mathbb{Z}, which sets are open? (a) \mathbb{Z} \setminus \{0, 5\}; (b) the even integers; (c) \{n \in \mathbb{Z} : |n| > 10\}; (d) the positive integers.
- On \mathbb{R} let d(x, y) = |e^x - e^y|, a metric. (a) Find \displaystyle B_d\left(0, \frac{1}{2}\right). (b) Find B_d(0, 2). (c) Find the largest r with B_d(0, r) \subseteq (-1, 1).
- Let U = \mathbb{R}^2 \setminus \{(0, 0)\} and P = (3, 4). For each of d_1, d_2 and d_\infty, find the largest r with B(P, r) \subseteq U. Why is U open in the standard topology?
- Is each set open in the standard topology on \mathbb{R}? In the finite-complement topology? (a) \displaystyle \mathbb{R} \setminus \{\sqrt{2}, \pi\}; (b) [0, \infty); (c) \mathbb{R} \setminus \mathbb{Q}; (d) \{x \in \mathbb{R} : x^2 \ne 4\}.
Comparing and building topologies
- On X = \{p, q, r\} let \mathcal{T}_1 = \{\varnothing, \{p\}, X\}, \mathcal{T}_2 = \{\varnothing, \{p\}, \{p, q\}, X\} and \mathcal{T}_3 = \{\varnothing, \{q\}, \{q, r\}, X\}. (a) Which pairs are comparable? (b) Is \mathcal{T}_2 \cup \mathcal{T}_3 a topology? (c) Find \mathcal{T}_2 \cap \mathcal{T}_3.
- In another watershed, water flows from u to w, from v to w, and from w into two channels, to x and to y. With the sealed sets of Example 5 as open sets: (a) list the open sets; (b) find the smallest open set containing u; (c) is \{w, x\} open?
Proofs
- Prove that a topology on X is discrete if and only if \{x\} is open for every x \in X.
- Let d, d' be metrics on X and c, C > 0 with c\, d(x, y) \le d'(x, y) \le C\, d(x, y) for all x, y. Prove that d and d' have the same metric topology.
- Let X have at least two points and fix x_0 \in X. Prove that \varnothing together with every U \subseteq X containing x_0 is a topology on X, and that it is not metrizable. For X = \mathbb{N} and x_0 = 0, is \{0, 3\} open? Is \{1, 2\}? (Here \mathbb{N} = \{0, 1, 2, \ldots\}.)
- Prove that a topology with exactly three open sets is not metrizable.
- Prove that if X is finite, the intersection of any nonempty family of open sets is open. Show that \displaystyle \bigcap_{n \ge 1} \left(2 - \frac{1}{n}, 5\right) is not open in the standard topology on \mathbb{R}.
- Prove that for every set X, the finite-complement topology is the coarsest topology on X in which X \setminus \{x\} is open for every x \in X.
Answers
- (a) A topology: the members are nested, so unions and intersections of two are members. (b) A topology: the only pair with neither containing the other is \{1, 2\}, \{3, 4\}, with union \{1, 2, 3, 4\} and intersection \varnothing. (c) Not a topology: \{1\} \cup \{2\} = \{1, 2\} is missing. (d) Not a topology: \varnothing is missing.
- It contains the intersections \{2\} and \{4\}, so every union of \{2\}, \{4\}, \{1, 2\}, \{2, 3, 4\}, \{4, 5\}. Any two of these five meet in \varnothing, \{2\}, \{4\} or one of the two, so by the distributive law the unions are closed under intersections of two, and they are the topology: \varnothing, \{2\}, \{4\}, \{2, 4\}, \{1, 2\}, \{4, 5\}, \{1, 2, 4\}, \{2, 4, 5\}, \{2, 3, 4\}, \{1, 2, 4, 5\}, \{1, 2, 3, 4\}, \{2, 3, 4, 5\} and X, 13 open sets.
- (a) A topology. Two rays meet in the ray to the smaller endpoint. A union containing \mathbb{R} is \mathbb{R}, one of \varnothing’s only is \varnothing, and otherwise a union of rays (-\infty, a_i) is \mathbb{R} if the a_i are unbounded above and (-\infty, s) with s = \sup a_i if not, since x < s exactly when x < a_i for some i. (b) Not a topology: \displaystyle \bigcup_{n \ge 1} \left(-\infty, -\frac{1}{n}\right] = (-\infty, 0) is not a member. (c) Not a topology: the rays (-\infty, a) over rational \displaystyle a < \sqrt{2} have union \displaystyle (-\infty, \sqrt{2}), as such rationals come arbitrarily close to \displaystyle \sqrt{2}, and \displaystyle \sqrt{2} \notin \mathbb{Q}.
- (a) Open: the complement \{0, 5\} is finite. (b) Not open: the odd integers are infinitely many. (c) Open: the complement \{-10, \ldots, 10\} has 21 elements. (d) Not open: the complement contains every integer n \le 0.
- (a) \displaystyle \frac{1}{2} < e^y < \frac{3}{2}, so \displaystyle B_d\left(0, \frac{1}{2}\right) = \left(-\ln 2, \ln \frac{3}{2}\right). (b) -1 < e^y < 3 holds whenever e^y < 3, so B_d(0, 2) = (-\infty, \ln 3). (c) For 0 < r < 1, B_d(0, r) = (\ln(1 - r), \ln(1 + r)), inside (-1, 1) when r \le 1 - e^{-1} and r \le e - 1. The largest is r = 1 - e^{-1} \approx 0.632.
- B(P, r) misses the origin exactly when d(P, (0, 0)) \ge r, so the largest r is that distance: d_1 = 7, d_2 = 5, d_\infty = 4. Each x \ne (0, 0) has B(x, \lVert x \rVert) \subseteq U, so U is open.
- (a) Open in both: the complement is two points (Theorem 4(b)). (b) Open in neither: no interval about 0 lies in it, and (-\infty, 0) is infinite. (c) Open in neither: every interval contains a rational, and \mathbb{Q} is infinite. (d) Open in both: the set is \mathbb{R} \setminus \{-2, 2\}.
- (a) Only \mathcal{T}_1 \subseteq \mathcal{T}_2, so \mathcal{T}_2 is strictly finer; \{p\} and \{q\} show the other pairs are not comparable. (b) Yes: in \{\varnothing, \{p\}, \{q\}, \{p, q\}, \{q, r\}, X\} the pairs that are not nested have unions \{p\} \cup \{q\} = \{p, q\} and \{p\} \cup \{q, r\} = \{p, q\} \cup \{q, r\} = X, and intersections \varnothing or \{q\}. (c) \{\varnothing, X\}.
- (a) A sealed set containing u, v or w contains w, x, y, while x, y are free: \varnothing, \{x\}, \{y\}, \{x, y\}, \{w, x, y\}, \{u, w, x, y\}, \{v, w, x, y\}, \{u, v, w, x, y\}, 8 in all. (b) \{u, w, x, y\}. (c) No: water from w reaches y.
- If the topology is discrete, every subset is open. Conversely, if every singleton is open, each nonempty U \subseteq X is \bigcup_{x \in U} \{x\}, open by (T2), and \varnothing is open by (T1). \square
- If d'(x, y) < cr then \displaystyle d(x, y) \le \frac{1}{c} d'(x, y) < r, so B_{d'}(x, cr) \subseteq B_d(x, r). If \displaystyle d(x, y) < \frac{r}{C} then d'(x, y) < r, so \displaystyle B_d\left(x, \frac{r}{C}\right) \subseteq B_{d'}(x, r). So a set contains a d-ball about each of its points exactly when it contains a d'-ball about each. \square
- \varnothing is in by definition, and X contains x_0. A union of members is \varnothing if all are empty and otherwise contains x_0; the intersection of two is \varnothing if either is empty and otherwise contains x_0. Pick y \ne x_0: every neighborhood of y contains x_0, so it meets every neighborhood of x_0, and Theorem 5 rules out a metric. On \mathbb{N}, \{0, 3\} is open and \{1, 2\} is not. \square
- The open sets are \varnothing, X and a third set A with \varnothing \ne A \ne X. Take x \in A and y \in X \setminus A. The only neighborhood of y is X, which contains x, so by Theorem 5 the topology is not metrizable. \square
- \mathcal{P}(X) is finite, so a nonempty family of open sets has finitely many distinct members, whose intersection is open by Theorem 1. Next, x lies in every \displaystyle \left(2 - \frac{1}{n}, 5\right) exactly when 2 \le x < 5, so the intersection is [2, 5), not open since every (2 - r, 2 + r) contains \displaystyle 2 - \frac{r}{2} \notin [2, 5). \square
- Each X \setminus \{x\} has finite complement, so it lies in the finite-complement topology \mathcal{T}_f. Let \mathcal{T} be a topology containing every X \setminus \{x\}, and U \in \mathcal{T}_f. If U = \varnothing or U = X, U \in \mathcal{T} by (T1). Otherwise F = X \setminus U is finite and nonempty, and U = \bigcap_{x \in F} (X \setminus \{x\}) is in \mathcal{T} by Theorem 1. So \mathcal{T}_f \subseteq \mathcal{T}. \square