Path-connectedness
A space in which any two points are joined by a continuous curve cannot be cut in two. Is every connected space of this kind?
Where you are 5 in Connectedness and compactness
Why this exists
Proving a space connected from the definition means ruling out every separation. It is often easier to join any two points by a continuous path. This lesson shows that such a space is connected (Connectedness) and finds where the two notions part: the topologist’s sine curve is connected, yet no path runs from one of its pieces to the other. Paths are also the raw material of Homotopy and the fundamental group.
Definition
Throughout, X is a topological space and I = [0, 1] has the standard topology.
Paths. A path in X from x to y is a continuous map f : I \to X with f(0) = x and f(1) = y. The constant path at x is c_x(s) = x. The reverse of f is \bar{f}(s) = f(1 - s). If f(1) = g(0), the product f \cdot g (“f first, then g”) is
Path-connected. X is path-connected if for all x, y \in X there is a path in X from x to y; the empty space qualifies. A subset A \subseteq X is path-connected when it is so in the subspace topology (Subspace and product topologies); by Theorem 3(c) there, a path in A is a path in X whose image lies in A.
Path components. The path component P_x of x \in X is the set of y \in X such that some path in X runs from x to y.
Locally path-connected. X is locally path-connected if for every x \in X and every open U containing x there is a path-connected open V with x \in V \subseteq U.
Convex sets. K \subseteq \mathbb{R}^n is convex if with any x, y \in K it contains the segment \{(1 - s)x + sy : s \in I\}.
Intuition
A path is a motion through X during the time interval I. It may stop, double back or cross itself; only its continuity matters. A separation U, V forbids a motion from U to V, since the moving point would have to jump. The converse fails when two pieces of a space come arbitrarily close with no route between them, as in Example 6. Unlike a path in a graph (Walks, paths, cycles and connected graphs), it is a map from an interval.
Results
Theorem 1 (joining). “Some path in X runs from x to y” is an equivalence relation on X: c_x is a path from x to x, \bar{f} is a path from f(1) to f(0), and f \cdot g is a path from f(0) to g(1) whenever f(1) = g(0). Its classes are the path components.
Theorem 2. Every path-connected space is connected.
Theorem 3 (continuous images). If X is path-connected and f : X \to Y is continuous, then f(X) is path-connected. Hence path-connectedness is a topological property.
Theorem 4 (unions and products). (a) If path-connected subsets A_j, j \in J, of X all contain a point p, then \bigcup_{j \in J} A_j is path-connected. (b) For nonempty X and Y, X \times Y is path-connected if and only if X and Y are.
Theorem 5 (convex sets). Every convex subset of \mathbb{R}^n is path-connected; so are \mathbb{R}^n, its open and closed balls, and every interval of \mathbb{R}.
Theorem 6 (path components). (a) The path components partition X, and P_x is the largest path-connected subset of X containing x. (b) P_x \subseteq C_x, the component of x; so each component is a union of path components.
Theorem 7 (locally path-connected spaces). If X is locally path-connected, each path component of X is open and closed, and equals the component containing it. In particular every connected open subset of \mathbb{R}^n is path-connected.
Why it is true
Theorem 1. Proof. c_x is constant, hence continuous (Theorem 3(a) of Continuous maps and homeomorphisms). \bar{f} = f \circ r with r(s) = 1 - s, continuous as |r(s) - r(t)| = |s - t|. The two formulas for f \cdot g agree at \displaystyle s = \frac{1}{2}, both giving f(1) = g(0), and on the closed sets \displaystyle [0, \frac{1}{2}] and \displaystyle [\frac{1}{2}, 1] they are composites of continuous maps; by the pasting lemma (Theorem 3(d) of Subspace and product topologies) f \cdot g is continuous. These facts are reflexivity, symmetry and transitivity, and the class of x is P_x by definition. \square
Theorem 2. Proof. The empty space is connected. Otherwise fix x \in X and, for each y \in X, a path f_y from x to y. As I is connected (Theorem 3 of Connectedness), f_y(I) is connected (Theorem 4 there) and contains x and y. So X = \bigcup_{y \in X} f_y(I) is a union of connected sets through x, connected by Theorem 5(a) there. \square
Theorem 3. Proof. Take f(x), f(y) \in f(X) and a path g in X from x to y. Then f \circ g is continuous (Theorem 3(d) of Continuous maps and homeomorphisms), runs from f(x) to f(y), and has image in f(X). A homeomorphism of X onto Y has image Y; apply this to it and to its inverse. \square
Theorem 4. Proof. (a) Let x \in A_j and y \in A_k. There are paths from x to p in A_j and from p to y in A_k, both paths in the union, and Theorem 1 joins x to y. (b) Let X, Y be path-connected and (x_1, y_1), (x_2, y_2) \in X \times Y. Take paths f from x_1 to x_2 and g from y_1 to y_2; then h(s) = (f(s), g(s)) is continuous by Theorem 5 of Subspace and product topologies and runs from (x_1, y_1) to (x_2, y_2). Conversely, \mathrm{pr}_1 is continuous and, as Y \ne \varnothing, maps X \times Y onto X, so X is path-connected by Theorem 3; likewise Y. \square
Theorem 5. Proof. For x, y in a convex K, f(s) = (1 - s)x + sy lies in K, runs from x to y, and is continuous since \lVert f(s) - f(t) \rVert = |s - t| \, \lVert y - x \rVert (Dot product, length and orthogonality in n dimensions). \mathbb{R}^n and the intervals are convex. For a ball about c of radius r, the triangle inequality bounds \lVert (1 - s)x + sy - c \rVert by (1 - s)\lVert x - c \rVert + s\lVert y - c \rVert, which is less than r (at most r) when x, y lie in the open (closed) ball. \square
Theorem 6. Proof. (a) The P_x partition X by Theorem 1. Let y, z \in P_x and let f be a path in X from y to z. For t \in I, s \mapsto f(ts) is a path from y to f(t), so f(t) \in P_y = P_x; thus f is a path in P_x, and P_x is path-connected. If A \ni x is path-connected, each a \in A is joined to x by a path in A, so A \subseteq P_x. (b) P_x is connected (Theorem 2) and contains x, so P_x \subseteq C_x by Theorem 8(a) of Connectedness. For y \in C_x, P_y \subseteq C_y = C_x, so C_x = \bigcup_{y \in C_x} P_y. \square
Theorem 7. Proof. Let y \in P_x. Some path-connected open V contains y, and V \subseteq P_y = P_x by Theorem 6(a). So P_x is open; its complement, the union of the other path components, is open for the same reason. Thus P_x is nonempty, open and closed in the connected subspace C_x (Theorem 6(b)), so P_x = C_x by Theorem 1 of Connectedness. For open W \subseteq \mathbb{R}^n, x \in W and U \ni x open in W, hence in \mathbb{R}^n, some ball B(x, r) \subseteq U, open and path-connected (Theorem 5). So W is locally path-connected, and if connected, one path component. \square
Worked examples
Routes for a stockroom robot
Andre and Leah program a robot that carries bins across a stockroom floor, the rectangle F = [0, 10] \times [0, 6] in meters. A rack fills Q_1 = [2, 8] \times [0, 4], and the robot may stand anywhere in X = F \setminus Q_1. (a) Give a path in X from A = (1, 1) to B = (9, 1). (b) Leah adds a rack Q_2 = [5, 6] \times [4, 6]. Show that no path in X' = F \setminus (Q_1 \cup Q_2) joins A to B, and find the path components of X'.
(a) Go north, east, then south, each leg in a third of the time: f(s) = (1, 1 + 12s) on \displaystyle [0, \frac{1}{3}], \displaystyle f(s) = (1 + 24(s - \frac{1}{3}), 5) on \displaystyle [\frac{1}{3}, \frac{2}{3}], and \displaystyle f(s) = (9, 5 - 12(s - \frac{2}{3})) on \displaystyle [\frac{2}{3}, 1]. Adjacent formulas agree at \displaystyle s = \frac{1}{3}, both giving (1, 5), and at \displaystyle s = \frac{2}{3}, both giving (9, 5); f(0) = A and f(1) = B. So f is continuous by the pasting lemma. The legs have x = 1 < 2, then y = 5 > 4, then x = 9 > 8, so none meets Q_1, and all stay in F.
(b) A point of F with 5 \le x \le 6 lies in Q_1 if y \le 4 and in Q_2 if y \ge 4, so no point of X' has 5 \le x \le 6. If g = (g_1, g_2) were a path in X' from a point with x < 5 to a point with x > 6, then g_1 would be continuous (Theorem 5 of Subspace and product topologies), and by the intermediate value theorem (The intermediate value theorem) \displaystyle g_1(c) = \frac{11}{2} for some c, putting g(c) outside X'. So A and B are not joined. The parts of X' with x < 5 and with x > 6 are
Each is the union of two convex sets meeting at (1, 5), respectively (9, 5), so it is path-connected (Theorems 5 and 4(a)). No path joins L to R, so the path components are L and R.
A path through a finite space
Let X = \{1, 2, 3, 4\} have the topology
Show that g : I \to X, with g(0) = 1, g(s) = 2 for \displaystyle 0 < s < \frac{1}{2}, \displaystyle g(\frac{1}{2}) = 3 and g(s) = 4 for \displaystyle \frac{1}{2} < s \le 1, is a path from 1 to 4, and that X is path-connected.
SolutionThe preimages of the open sets, in the order listed, are \varnothing, \displaystyle (0, \frac{1}{2}), \displaystyle (\frac{1}{2}, 1], \displaystyle (0, \frac{1}{2}) \cup (\frac{1}{2}, 1], \displaystyle [0, \frac{1}{2}), \displaystyle I \setminus \{\frac{1}{2}\}, (0, 1] and I. Each is open in I; for instance \displaystyle (\frac{1}{2}, 1] = I \cap (\frac{1}{2}, 2) and \displaystyle [0, \frac{1}{2}) = I \cap (-1, \frac{1}{2}). So g is continuous. For each t, s \mapsto g(ts) is a path from 1 to g(t), and g takes every value, so every point is joined to 1, and any two points to each other by Theorem 1. A path in \mathbb{R} with finite image is constant; here g can jump because every open set containing 3 contains 2 and 4.
Spheres
Prove that \mathbb{R}^{n+1} \setminus \{0\} and the sphere S^n = \{x \in \mathbb{R}^{n+1} : \lVert x \rVert = 1\} are path-connected for n \ge 1, and that S^0 is not.
SolutionProof. Let x, y \ne 0. If (1 - s)x + sy = 0 for some s \in I, then s \ne 0, 1 and \displaystyle y = -\frac{1 - s}{s} x lies on the line \mathbb{R}x. Otherwise the segment misses 0. If y \in \mathbb{R}x, pick z \notin \mathbb{R}x, possible as n + 1 \ge 2; the segments from x to z and from z to y miss 0, or z would lie on \mathbb{R}x = \mathbb{R}y. So one segment, or the product of two, joins x to y in \mathbb{R}^{n+1} \setminus \{0\}. The norm is continuous, as |\lVert x \rVert - \lVert y \rVert| \le \lVert x - y \rVert, and nonzero there, so \displaystyle x \mapsto \frac{x}{\lVert x \rVert} is continuous (Theorems 5 and 6(b) of Subspace and product topologies), and it maps onto S^n. By Theorem 3, S^n is path-connected. S^0 = \{-1, 1\} is not an interval, so not connected, nor path-connected (Theorem 2). \square
Rotations and reflections
Identify a real 2 \times 2 matrix with the point of \mathbb{R}^4 formed by its entries, and let O(2) = \{A : A^{\mathsf{T}} A = I_2\}, where I_2 is the identity matrix. (a) Show that every A \in O(2) is R_\theta or F_\theta for some \theta \in \mathbb{R}, where
(b) Give a path in O(2) from R_{\pi/2} to -I_2. (c) Find the path components of O(2).
Solution(a) A^{\mathsf{T}} A = I_2 says the columns of A are orthonormal. The first is a unit vector, (\cos\theta, \sin\theta) for some \theta (The unit circle). The vectors orthogonal to it form the line through (-\sin\theta, \cos\theta), and unit length leaves \pm(-\sin\theta, \cos\theta): these give R_\theta and F_\theta. Conversely both satisfy A^{\mathsf{T}} A = I_2, as \cos^2\theta + \sin^2\theta = 1; and \det R_\theta = 1, \det F_\theta = -1.
(b) \theta \mapsto R_\theta is continuous from \mathbb{R} to \mathbb{R}^4, each entry being continuous (Theorem 5 of Subspace and product topologies), and R_\pi = -I_2. So s \mapsto R_{\pi/2 + \pi s/2} is a path from R_{\pi/2} to -I_2.
(c) O^{+} = \{R_\theta : \theta \in \mathbb{R}\} is a continuous image of \mathbb{R}, so path-connected (Theorems 5 and 3), and so is O^{-} = \{F_\theta : \theta \in \mathbb{R}\}. The determinant, a polynomial in the entries, is continuous with values \pm 1 on O(2). A path g in O(2) from O^{+} to O^{-} would make \det \circ g map I onto a connected subset of \{-1, 1\} holding both values, contradicting Theorem 3 of Connectedness. So the path components are O^{+} and O^{-}.
The plane minus a countable set
Let C \subseteq \mathbb{R}^2 be countable (Countable sets). Prove that \mathbb{R}^2 \setminus C is path-connected.
Proof. Let p \ne q in \mathbb{R}^2 \setminus C, and let L be the perpendicular bisector of the segment from p to q, with closed half-planes H_p \ni p and H_q \ni q. For z \in L let \gamma_z be the product of the segment paths from p to z and from z to q. The segment [p, z] lies in H_p, a convex set containing p and z, and meets L only at z, as p \notin L; likewise [z, q] \subseteq H_q. Take z \ne z' in L. The lines through p and z and through p and z' differ, as the first meets L only at z, so [p, z] \cap [p, z'] = \{p\}; likewise [z, q] \cap [z', q] = \{q\}. And [p, z] \cap [z', q] \subseteq H_p \cap H_q = L, so it lies in \{z\} \cap \{z'\} = \varnothing. Hence the sets \gamma_z(I) \setminus \{p, q\}, z \in L, are pairwise disjoint. Each point of C lies in at most one of them, so at most countably many \gamma_z meet C, while L, a copy of \mathbb{R}, is uncountable (Countable and uncountable sets). Some \gamma_z misses C, and as p, q \notin C, it is a path in \mathbb{R}^2 \setminus C. \square
The topologist’s sine curve
Let \displaystyle S = \{(x, \sin \frac{1}{x}) : 0 < x \le 1\}, V = \{0\} \times [-1, 1] and T = S \cup V \subseteq \mathbb{R}^2. Prove that T is connected but not path-connected, with path components S and V.
Proof. S is the image of (0, 1] under the continuous \displaystyle x \mapsto (x, \sin \frac{1}{x}), so path-connected (Theorems 5 and 3), and V is convex. For |c| \le 1 and positive integers k, let \displaystyle x_k = \frac{1}{\arcsin c + 2k\pi}. The denominator is at least \displaystyle 2k\pi - \frac{\pi}{2} > 1, so (x_k, c) \in S, and (x_k, c) \to (0, c). So S \subseteq T \subseteq \overline{S}, and T is connected by Theorem 6 of Connectedness.
Suppose f = (f_1, f_2) is a path in T from a point of V to a point of S; reversing handles the other direction. V = T \cap (\{0\} \times \mathbb{R}) is closed in T, so f^{-1}(V) is closed in I and has a largest element t_0, with t_0 < 1. For t > t_0, f(t) \in S, so f_1(t) > 0 and \displaystyle f_2(t) = \sin \frac{1}{f_1(t)}; and f_1(t_0) = 0. By continuity of f_2, some \delta > 0 has |f_2(t) - f_2(t_0)| < 1 for t \in I with |t - t_0| < \delta. Pick t_1 \in I with t_0 < t_1 < t_0 + \delta, so a = f_1(t_1) > 0, and a positive integer k with \displaystyle u = \frac{1}{2k\pi + \pi/2} < a; then also \displaystyle v = \frac{1}{2k\pi + 3\pi/2} < a. The intermediate value theorem for f_1 on [t_0, t_1] gives t, t' in (t_0, t_1) with f_1(t) = u and f_1(t') = v, so f_2(t) = 1 and f_2(t') = -1. Both lie within 1 of f_2(t_0), so 2 = |f_2(t) - f_2(t')| < 2, a contradiction. No path joins V to S, and as both are path-connected, they are the path components. \square
Since T = \overline{S} (Exercise 13), the closure of a path-connected set need not be path-connected, and the path component S is not closed in T.
Boundaries
- Path-connected implies connected, not conversely: Example 6 and the deleted comb (Exercise 11). The two agree for locally path-connected spaces (Theorem 7), hence for open subsets of \mathbb{R}^n, and also for subsets of \mathbb{R} (Exercise 9) and finite spaces (Exercise 8).
- A path is a map, not its image. The image need not be a simple arc: Peano constructed a continuous map from I onto the square [0, 1]^2.
Where this goes next
Homotopy deforms paths with endpoints fixed, and loops, paths that end where they start, build the fundamental group.
Exercises
Paths
- Let A = \{(x, y) : 1 \le x^2 + y^2 \le 4\}. Give a path in A from (2, 0) to (-1, 0) and check that it stays in A.
- Let f(s) = (s, s^2) and g(s) = (1 + s, 1 - s) in \mathbb{R}^2. (a) Write f \cdot g and \bar{g}. (b) Find \displaystyle (f \cdot g)(\frac{1}{4}) and \displaystyle (f \cdot g)(\frac{3}{4}).
- Let X = \{1, 2, 3, 4, 5\} have the topology with subbasis \{1\}, \{1, 2\}, \{3\}, \{3, 4\}, \{1, 3, 5\}. (a) Show that h, with h(0) = 2, h = 1 on \displaystyle (0, \frac{1}{2}), \displaystyle h(\frac{1}{2}) = 5, h = 3 on \displaystyle (\frac{1}{2}, 1) and h(1) = 4, is a path. (b) Find the path components of the subspace \{1, 2, 3, 4\}.
- In a second stockroom, also F = [0, 10] \times [0, 6], Andre places racks P = [2, 4] \times [0, 5] and Q = [6, 8] \times [1, 6]. (a) Give a polygonal path for the robot from (1, 1) to (9, 5). (b) Leah adds the rack [4, 6] \times [0, 1]. Show that no path joins the two points now.
Path components
- How many path components has \{(x, y) \in \mathbb{R}^2 : x^2 \ne y^2\}?
- Let N = \{(x, y) : y^2 = x^2(x + 1)\}. How many path components have (a) N and (b) N \setminus \{(0, 0)\}?
- Which are path-connected? (a) The union of the closed disks of radius 1 about (0, 0) and (2, 0). (b) The union of the open disks with the same centers and radius. (c) \displaystyle \Gamma = \{(x, \sin \frac{1}{x}) : x \ne 0\} \cup \{(0, 0)\}. (d) S^1 \times \mathbb{R}.
Proofs
- Prove that a finite space is connected if and only if it is path-connected.
- Prove that a subset of \mathbb{R} is path-connected if and only if it is an interval.
- Prove that any two points of a connected open W \subseteq \mathbb{R}^n are joined by a polygonal path in W, a product of finitely many segment paths.
- The deleted comb D is the union of I \times \{0\}, the segments \displaystyle \{\frac{1}{n}\} \times I for n \ge 1, and the point (0, 1). Prove that D is connected but not path-connected.
- Prove that a continuous f : X \to Y maps each path component of X into a path component of Y, and that a homeomorphism matches the path components of X and Y one to one.
- Prove that the space T of Example 6 is the closure of S in \mathbb{R}^2.
Answers
- f(s) = ((2 - s)\cos \pi s, (2 - s)\sin \pi s). Its coordinates are continuous, f(0) = (2, 0), f(1) = (-1, 0), and \lVert f(s) \rVert = 2 - s \in [1, 2], so f(s) \in A.
- (a) (f \cdot g)(s) = (2s, 4s^2) for \displaystyle 0 \le s \le \frac{1}{2} and (2s, 2 - 2s) for \displaystyle \frac{1}{2} \le s \le 1, both (1, 1) at \displaystyle s = \frac{1}{2}; \bar{g}(s) = (2 - s, s). (b) \displaystyle (\frac{1}{2}, \frac{1}{4}) and \displaystyle (\frac{3}{2}, \frac{1}{2}).
- (a) The subbasis sets have preimages \displaystyle (0, \frac{1}{2}), \displaystyle [0, \frac{1}{2}), \displaystyle (\frac{1}{2}, 1), \displaystyle (\frac{1}{2}, 1], (0, 1), all open in I, so h is continuous (Theorem 2(e) of Continuous maps and homeomorphisms), a path from 2 to 4. (b) \{1, 2\} and \{3, 4\}. They are open in the subspace and complementary, so no path crosses (Theorem 2); the path equal to 2 at 0 and 1 elsewhere joins 2 to 1, and likewise 4 to 3.
- (a) Segments from (1, 1) through (1, 5.5), (5, 5.5), (5, 0.5), (9, 0.5) to (9, 5). The legs have x = 1 < 2; y = 5.5 > 5 with x \le 5 < 6; x = 5, between the racks; y = 0.5 < 1 with x \ge 5 > 4; x = 9 > 8. (b) The line x = 6 is now blocked: (6, y) lies in the new rack for y \le 1 and in Q for y \ge 1. A path’s x-coordinate runs continuously from 1 to 9 and takes the value 6.
- 4: the open wedges \{|y| < x\}, \{|y| < -x\}, \{|x| < y\}, \{|x| < -y\}. Each is an intersection of two open half-planes, convex. They are disjoint and open and cover the set, so a path, with connected image, stays in one.
- (a) 1: t \mapsto (t^2 - 1, t(t^2 - 1)) maps \mathbb{R} onto N, as for x \ne 0 the point is reached at \displaystyle t = \frac{y}{x} (then t^2 = x + 1), and (0, 0) at t = 1. (b) 3: the images of t < -1, -1 < t < 1, t > 1, which lie in the disjoint open sets \{x > 0, y < 0\}, \{x < 0\}, \{x > 0, y > 0\}; only t = \pm 1 map to (0, 0).
- (a) Yes: two convex sets meeting at (1, 0), Theorem 4(a). (b) No: a common point would be less than 1 from both centers, which are 2 apart; the disks are open, disjoint and cover the union, so they are its two path components. (c) No, but connected. The closure of each half contains (0, 0), as \displaystyle (\frac{1}{k\pi}, 0) \in \Gamma, so each half with (0, 0) is connected, and so is their union. No path runs from (0, 0) to (1, \sin 1): after its last time at (0, 0) it stays in \{x > 0\}, which with \{x < 0\} separates \Gamma \setminus \{(0, 0)\}; then argue as in Example 6. (d) Yes, Theorem 4(b) with Example 3.
- Path-connected spaces are connected (Theorem 2). For x in a finite X, the intersection M_x of the open sets containing x is open. For z \in M_x, let q(0) = x and q(t) = z for t > 0. For open U, q^{-1}(U) is I if x \in U (then z \in M_x \subseteq U), (0, 1] if z \in U and x \notin U, and \varnothing otherwise; so q is a path in M_x from x to z. Thus M_x is path-connected (Theorem 1) and lies in every open set containing x, so X is locally path-connected, and Theorem 7 makes a connected X path-connected. \square
- An interval is convex, so path-connected (Theorem 5). A path-connected subset is connected (Theorem 2), so an interval by Theorem 3 of Connectedness. \square
- Fix x \in W and let P be the set of points joined to x by a polygonal path in W. For y \in W take B(y, r) \subseteq W; it is convex, so each of its points is joined to y by a segment in it. If y \in P, adding a segment gives B(y, r) \subseteq P; if y \notin P, the ball misses P, or y would be reached through it. So P and W \setminus P are open, x \in P, and P = W as W is connected. \square
- Let p = (0, 1). D \setminus \{p\} is the base segment with the teeth \displaystyle \{\frac{1}{n}\} \times I, each meeting it, so path-connected (Theorems 5 and 4(a)); and \displaystyle (\frac{1}{n}, 1) \to p, so D is connected by Theorem 6 of Connectedness. Let f be a path in D with f(0) = p, and Z = f^{-1}(\{p\}), closed. For t \in Z, continuity gives an interval J \ni t, open in I, with \displaystyle f(J) \subseteq B(p, \frac{1}{2}), and f(J) is connected. \displaystyle D \cap B(p, \frac{1}{2}) holds no base point, so besides p only points \displaystyle (\frac{1}{n}, y). If f(J) held one, any c with \displaystyle \frac{1}{n + 1} < c < \frac{1}{n} makes \{x < c\} and \{x > c\} separate f(J). So f(J) = \{p\}, Z is open, and Z = I: every path from p is constant, and none reaches (1, 0). \square
- f(P_x) is path-connected (Theorems 6(a) and 3) and contains f(x), so f(P_x) \subseteq P_{f(x)} by Theorem 6(a). For a homeomorphism h, also h^{-1}(P_{h(x)}) \subseteq P_x, so h(P_x) = P_{h(x)}, and P_x \mapsto h(P_x) is a bijection with inverse given by h^{-1}. \square
- Example 6 shows T \subseteq \overline{S}. Let (a, b) \notin T. If a < 0 or a > 1, the open half-plane \{x < 0\} or \{x > 1\} contains it and misses S. If a = 0, then |b| > 1, and \{|y| > 1\} misses S. If 0 < a \le 1, then \displaystyle b \ne \sin \frac{1}{a}, and \displaystyle \{(x, y) : x > 0, \ y \ne \sin \frac{1}{x}\} is open (a preimage under a continuous map on \{x > 0\}), contains (a, b) and misses S. So \overline{S} \subseteq T. \square