math.wiki
Topology › Connectedness and compactness › Compactness in products and the Heine-Borel theorem
Sign in

Compactness in products and the Heine-Borel theorem

Every closed interval [a, b] is compact. Is a closed square, a solid ball or a set of matrices compact, and is there a test as quick as “closed and bounded” that decides it?

Where you are 5 in Connectedness and compactness

Why this exists

Compactness and open covers proved that a set of real numbers is compact exactly when it is closed and bounded. The square [0, 1] \times [0, 1] is a product of two intervals, so the first question is whether a product of compact spaces is compact. The tube lemma answers it with open sets alone, and then every closed bounded set of \mathbb{R}^n sits inside a compact cube. Metric spaces, a first look reached the same test through sequences; the argument here is the one that extends to products of spaces with no metric.

Compactness also gives covers a grain: on a compact metric space every open cover has a Lebesgue number, a size below which every set fits inside one member. That number lets a path be chopped into pieces small enough to handle one at a time, the step The fundamental group of the circle is built on.

Definition

Throughout, X, Y and X_1, \dots, X_n are topological spaces. A product X_1 \times \cdots \times X_n carries the product topology, whose basis is the sets U_1 \times \cdots \times U_n with each U_i open in X_i (Subspace and product topologies, built on the Cartesian product). \mathbb{R}^n carries its standard topology, which is both the product topology (Theorem 6(a) of Subspace and product topologies) and the topology of the metric d(x, y) = \lVert x - y \rVert (Dot product, length and orthogonality in n dimensions).

Slices and tubes. For x_0 \in X, the slice over x_0 is \{x_0\} \times Y \subseteq X \times Y. A tube about it is a set W \times Y with W open in X and x_0 \in W.

Bounded sets and diameter. A subset A of a metric space (X, d) is bounded if A \subseteq B(p, R) for some p \in X and R > 0; in \mathbb{R}^n we may take p = 0, since B(p, R) \subseteq B(0, R + \lVert p \rVert). The diameter of a nonempty bounded A is

\operatorname{diam} A = \sup\{d(a, b) : a, b \in A\}.

Distance to a set. For x \in X and nonempty A \subseteq X, d(x, A) = \inf\{d(x, a) : a \in A\}.

Lebesgue number. Let \mathcal{U} be an open cover of a metric space X. A number \delta > 0 is a Lebesgue number of \mathcal{U} if every nonempty bounded A \subseteq X with \operatorname{diam} A < \delta lies inside a single member of \mathcal{U}. Every smaller positive number is then one too.

Intuition

Let N be an open set of X \times Y containing the slice \{x_0\} \times Y. Around each point (x_0, y) of the slice, N holds a basic box, wide at some heights and narrow at others. If infinitely many boxes are needed, their widths can shrink to 0, and no single band about x_0 fits inside N all the way up. When Y is compact, finitely many boxes suffice, the narrowest still has positive width, and a whole tube W \times Y lies inside N (Figure 1).

N
x_0
W \times Y
Figure 1The tube lemma: N may pinch, but over a compact Y it never pinches down to the slice, so a whole tube W \times Y fits inside.

The tube lemma lets finite subcovers found slice by slice be assembled into one finite subcover of X \times Y. The Lebesgue number is the same finiteness seen through a metric: finitely many members leave a smallest margin that works everywhere at once.

Results

Theorem 1 (tube lemma). Let Y be compact, x_0 \in X, and N an open set of X \times Y containing \{x_0\} \times Y. Then there is an open W \subseteq X with x_0 \in W and W \times Y \subseteq N.

Theorem 2 (finite products). If X and Y are compact, then X \times Y is compact. Hence if X_1, \dots, X_n are compact, so is X_1 \times \cdots \times X_n.

Theorem 3 (Tychonoff’s theorem). For any family of spaces X_\alpha, \alpha \in J, give \prod_{\alpha \in J} X_\alpha the product topology, with basis the products \prod U_\alpha in which every U_\alpha is open in X_\alpha and U_\alpha = X_\alpha for all but finitely many \alpha. If every X_\alpha is compact, the product is compact.

Theorem 4 (Heine-Borel theorem). A subset A \subseteq \mathbb{R}^n is compact if and only if it is closed in \mathbb{R}^n and bounded.

Theorem 5 (distance to a set). Let (X, d) be a metric space and A \subseteq X nonempty. For all x, y \in X, |d(x, A) - d(y, A)| \le d(x, y), so x \mapsto d(x, A) is continuous; and d(x, A) = 0 if and only if x \in \overline{A}.

Theorem 6 (Lebesgue number lemma). Every open cover of a compact metric space has a Lebesgue number.

Theorem 7 (uniform continuity). Let (X, d) be a compact metric space, (Y, e) a metric space and f : X \to Y continuous. For every \varepsilon > 0 there is \delta > 0 with e(f(a), f(b)) < \varepsilon whenever d(a, b) < \delta.

Theorem 7 extends Theorem 4 of Uniform continuity from compact subsets of \mathbb{R} to every compact metric space.

Why it is true

Theorem 1. Proof. For each y \in Y the point (x_0, y) lies in the open set N, so some basis element satisfies (x_0, y) \in U_y \times V_y \subseteq N. The open sets V_y cover Y, so finitely many V_{y_1}, \dots, V_{y_k} do. Put W = U_{y_1} \cap \cdots \cap U_{y_k}, open and containing x_0 (if Y = \varnothing, then k = 0 and W = X). For (x, y) \in W \times Y, choose i with y \in V_{y_i}; since x \in W \subseteq U_{y_i}, (x, y) \in U_{y_i} \times V_{y_i} \subseteq N. \square

Theorem 2. Proof. Let \mathcal{A} be an open cover of X \times Y and fix x_0 \in X. The map y \mapsto (x_0, y) is continuous, its coordinates being a constant and \mathrm{id}_Y (Theorem 5 of Subspace and product topologies), so the slice \{x_0\} \times Y is compact (Theorem 5 of Compactness and open covers). By Theorem 1 of Compactness and open covers, finitely many members of \mathcal{A} cover the slice; let N be their union. Theorem 1 gives a tube W_{x_0} \times Y \subseteq N, covered by those finitely many members. Finitely many W_{x_1}, \dots, W_{x_m} cover X, so the tubes W_{x_j} \times Y cover X \times Y, and their finitely many finite families together form a finite subcover.

For n factors, use induction. The bijection X_1 \times \cdots \times X_n \to (X_1 \times \cdots \times X_{n - 1}) \times X_n that regroups the coordinates sends the basis element U_1 \times \cdots \times U_n to (U_1 \times \cdots \times U_{n - 1}) \times U_n, and these form a basis of the right side by Theorem 4(b) of Subspace and product topologies. A bijection carrying a basis onto a basis carries open sets onto open sets both ways, so it is a homeomorphism, and the two-factor case finishes. \square

Theorem 3 is beyond this lesson. Its proof needs Zorn’s lemma, a form of the axiom of choice, and is done in Munkres, Topology, section 37. Exercise 10 proves a converse for two factors.

Theorem 4. Proof. Let A be closed and bounded, say \lVert x \rVert < R for every x \in A. Each coordinate satisfies |x_i| \le \lVert x \rVert < R, so A \subseteq Q = [-R, R]^n. The interval [-R, R] is compact (Theorem 9 of Compactness and open covers), so Q with the product topology is compact (Theorem 2). That product topology is the subspace topology Q inherits from \mathbb{R}^n (Theorems 4(d) and 6(a) of Subspace and product topologies). Since A = A \cap Q is closed in Q, it is compact by Theorem 3 of Compactness and open covers.

Conversely, let A be compact. \mathbb{R}^n is a metric space, hence Hausdorff (Theorem 1 of Hausdorff spaces and separation), so A is closed (Theorem 4 of Compactness and open covers). The open balls B(0, m), m \ge 1, cover \mathbb{R}^n and increase with m, so one ball B(0, M) contains A. \square

Theorem 5. Proof. For each a \in A, d(x, A) \le d(x, a) \le d(x, y) + d(y, a). Taking the infimum over a gives d(x, A) \le d(x, y) + d(y, A), and exchanging x and y gives the inequality; \delta = \varepsilon then works in Theorem 4 of Continuous maps and homeomorphisms. Next, d(x, A) = 0 exactly when every ball B(x, r) meets A, and since the balls form a basis this happens exactly when x \in \overline{A} (Theorem 3 of Closed sets, closure and limit points). \square

Theorem 6. Proof. Let \mathcal{U} cover the compact metric space X \ne \varnothing. If X itself is a member, every \delta > 0 works. Otherwise choose a finite subcover U_1, \dots, U_m and put C_i = X \setminus U_i, closed and nonempty. Define

f(x) = \frac{1}{m} \sum_{i = 1}^{m} d(x, C_i),

continuous by Theorem 5 and Theorem 6(b) of Subspace and product topologies. Each x lies in some U_j, so x \notin C_j = \overline{C_j} and d(x, C_j) > 0 by Theorem 5; hence f(x) > 0. By Theorem 7 of Compactness and open covers, f has a least value \delta > 0.

Let A be nonempty with \operatorname{diam} A < \delta and pick a \in A. An average is at most its largest term, so some j has d(a, C_j) \ge f(a) \ge \delta. Every b \in A has d(a, b) \le \operatorname{diam} A < \delta \le d(a, C_j), so b \notin C_j, that is b \in U_j. So A \subseteq U_j. \square

Theorem 7. Proof. Let \varepsilon > 0. The sets \displaystyle f^{-1}\big(B(y, \frac{\varepsilon}{2})\big), y \in Y, are open and cover X; let \delta be a Lebesgue number (Theorem 6). If d(a, b) < \delta, the set \{a, b\} has diameter less than \delta, so both points lie in one \displaystyle f^{-1}\big(B(y, \frac{\varepsilon}{2})\big), and e(f(a), f(b)) \le e(f(a), y) + e(y, f(b)) < \varepsilon. \square

Worked examples

Example 1

Three sets in space

Decide which sets are compact. (a) A = \{(x, y, z) \in \mathbb{R}^3 : x^2 + y^2 \le z \le 4\}. (b) C = \{(x, y) \in \mathbb{R}^2 : y^2 = x^3(1 - x)\}. (c) H = \{(x, y) \in \mathbb{R}^2 : xy = 1\}.

Solution

Each test is Theorem 4. Polynomials are continuous on \mathbb{R}^n (Theorem 6(b) of Subspace and product topologies), and preimages of closed sets under continuous maps are closed (Continuous maps and homeomorphisms).

(a) A is the intersection of the preimage of [0, \infty) under z - x^2 - y^2 and the preimage of (-\infty, 4] under z, so it is closed. On A, 0 \le x^2 + y^2 \le z \le 4, so \lVert p \rVert^2 \le 4 + 4^2 = 20. So A lies in the closed ball of radius \displaystyle 2\sqrt{5} and is compact.

(b) C is the preimage of \{0\} under y^2 - x^3(1 - x), so it is closed. A point of C has x^3(1 - x) = y^2 \ge 0, which forces 0 \le x \le 1. The derivative of x^3 - x^4 is x^2(3 - 4x), so on [0, 1] its largest value is at \displaystyle x = \frac{3}{4}, namely \displaystyle \frac{27}{256}. So \displaystyle y^2 \le \frac{27}{256}, and C is bounded and compact.

(c) H is closed, as the preimage of \{1\} under xy, but it holds \displaystyle (m, \frac{1}{m}) with norm greater than m for every m \ge 1, so it is not bounded and not compact.

Example 2

Misters over a propagation bench

Naomi runs a propagation bench B = [0, 5] \times [0, 2], measured in meters. Six misters each wet an open rectangle U_i \times V_j, where

\begin{aligned} U_1 &= (-1, 2), \; U_2 = (1.5, 4), \; U_3 = (3.2, 6), \\ V_1 &= (-1, 1.2), \; V_2 = (0.6, 3). \end{aligned}

She wants every seed tray, of any shape, set anywhere on the bench to sit wholly in one mister’s zone. Find the largest \delta such that every tray of diameter less than \delta does.

0
1.5
2
3.2
4
5
0.6
1.2
2
Figure 2Example 2: the shaded strips are where two zones overlap. The narrowest is 1.5 < x < 2, and the dots at its two ends lie in no common zone.
Solution

The U_i cover [0, 5] and the V_j cover [0, 2], so the sets (U_i \times V_j) \cap B form an open cover of the compact metric space B (Theorem 4), and Theorem 6 promises a Lebesgue number. We find the largest one: \displaystyle \delta = \frac{1}{2}.

Let S \subseteq B be nonempty with \displaystyle \operatorname{diam} S < \frac{1}{2}. Since |x - x'| \le \lVert p - p' \rVert, the first coordinates of S form a set S_1 \subseteq [0, 5] with \displaystyle \sup S_1 - \inf S_1 < \frac{1}{2}. If \sup S_1 < 2, then S_1 \subseteq U_1. Otherwise \inf S_1 > 1.5, and S_1 \subseteq U_2 if \sup S_1 < 4. Otherwise \inf S_1 > 3.5 and \sup S_1 \le 5, so S_1 \subseteq U_3. Likewise for the second coordinates S_2: if \sup S_2 < 1.2, S_2 \subseteq V_1; otherwise \inf S_2 > 0.7 and S_2 \subseteq V_2. So S \subseteq U_i \times V_j for suitable i, j.

No larger number works. The points (1.5, 0) and (2, 0) are \displaystyle \frac{1}{2} apart, the only U_i containing 1.5 is U_1, and U_1 misses 2; so a round tray of diameter exactly \displaystyle \frac{1}{2} centered at (1.75, 0.3) fits in no zone. The overlaps are \displaystyle \frac{1}{2} and 0.8 across and 0.6 up, and the smallest sets the answer.

Example 3

No tube without compactness

Let N = \{(x, y) \in \mathbb{R}^2 : |xy| < 1\}. Prove that N is open and contains the slice \{0\} \times \mathbb{R}, but contains no tube W \times \mathbb{R} about it. Then show that \displaystyle (-\frac{1}{c}, \frac{1}{c}) \times [-c, c] \subseteq N for every c > 0.

N
W \times \mathbb{R}
Figure 3Example 3: N = \{(x, y) : |xy| < 1\} contains the whole axis \{0\} \times \mathbb{R}, yet every tube about the axis leaves it.
Solution

Proof. N is the preimage of the open set (-1, 1) under the continuous xy, so it is open, and |0 \cdot y| = 0 < 1 puts the slice inside. Let W be open with 0 \in W. It contains an interval (-r, r) with r > 0, so \displaystyle (\frac{r}{2}, \frac{4}{r}) \in W \times \mathbb{R}, while \displaystyle \frac{r}{2} \cdot \frac{4}{r} = 2, so that point is not in N. No tube lies in N.

Over the compact factor [-c, c] a tube exists, as Theorem 1 says it must: if \displaystyle |x| < \frac{1}{c} and |y| \le c, then \displaystyle |xy| < \frac{1}{c} \cdot c = 1. \square

Example 4

A compact set keeps its distance

Let (X, d) be a metric space, K \subseteq X nonempty and compact, and C \subseteq X nonempty, closed and disjoint from K. Prove that some k_0 \in K has d(k_0, C) = \inf\{d(k, c) : k \in K, c \in C\} > 0. Show that this fails for two disjoint closed sets, and find the distance from the closed disk D = \{(x, y) : x^2 + y^2 \le 1\} to the hyperbola \{(x, y) : xy = 4\}.

Solution

Proof. By Theorem 5, g(x) = d(x, C) is continuous, and g(k) > 0 for k \in K because k \notin C = \overline{C}. By Theorem 7 of Compactness and open covers, g has a least value on K at some k_0, and g(k_0) > 0. For all k \in K and c \in C, d(k, c) \ge g(k) \ge g(k_0), and the infimum of d(k_0, c) over c is g(k_0), so the infimum over both is g(k_0). \square

Without compactness: the branch \displaystyle \{(x, \frac{1}{x}) : x > 0\} and the x-axis are closed and disjoint, but \displaystyle (x, \frac{1}{x}) and (x, 0) are \displaystyle \frac{1}{x} apart, which tends to 0.

For the disk: on the hyperbola x^2 + y^2 \ge 2|xy| = 8, with equality exactly at (2, 2) and (-2, -2). For q \in D and p on the hyperbola, \displaystyle \lVert p - q \rVert \ge \lVert p \rVert - \lVert q \rVert \ge 2\sqrt{2} - 1, and \displaystyle q = (\frac{\sqrt{2}}{2}, \frac{\sqrt{2}}{2}), p = (2, 2) reach it. The distance is \displaystyle 2\sqrt{2} - 1.

Example 5

Chopping a path on the circle

Let f : I \to S^1, f(s) = (\cos 3\pi s, \sin 3\pi s), and cover S^1 by U = S^1 \setminus \{(-1, 0)\} and V = S^1 \setminus \{(1, 0)\}. Find the largest Lebesgue number of the cover \{f^{-1}(U), f^{-1}(V)\} of I, and the least n for which each piece \displaystyle [\frac{k - 1}{n}, \frac{k}{n}], 1 \le k \le n, is carried by f into U or into V.

Solution

f(s) = (-1, 0) when 3s is an odd integer and f(s) = (1, 0) when 3s is even, so

f^{-1}(U) = I \setminus \{\tfrac{1}{3}, 1\}, \qquad f^{-1}(V) = I \setminus \{0, \tfrac{2}{3}\}.

A set lies in neither exactly when it holds a point of \displaystyle \{\frac{1}{3}, 1\} and a point of \displaystyle \{0, \frac{2}{3}\}. The four distances between such points are \displaystyle \frac{1}{3}, \displaystyle \frac{1}{3}, 1 and \displaystyle \frac{1}{3}, so every set of diameter less than \displaystyle \frac{1}{3} lies in one of the two, while \displaystyle \{0, \frac{1}{3}\} lies in neither. The largest Lebesgue number is \displaystyle \frac{1}{3}.

A piece has diameter \displaystyle \frac{1}{n}, so n \ge 4 works. For n \le 3 the first piece \displaystyle [0, \frac{1}{n}] contains both 0 and \displaystyle \frac{1}{3}. The least n is 4. On each piece f misses a point of the circle, which is how The fundamental group of the circle lifts paths, piece by piece.

Boundaries

  • Closed and bounded is a fact about \mathbb{R}^n and its metric. For the metric \min\{\lVert x - y \rVert, 1\} on \mathbb{R}^n the whole space is closed and bounded but not compact (Exercise 14). An infinite set with the discrete metric is closed and bounded, and its singletons form an open cover with no finite subcover.
  • The tube lemma needs the factor compact (Example 3), and so does the Lebesgue number: Exercise 7 gives an open cover of (0, 1) with none.
  • A Lebesgue number is enough, not necessary. A coarser chopping may fit by luck (Exercise 8).
  • Infinite products. Tychonoff’s theorem needs the product topology. In the box topology, where every product of open sets is open, infinitely many copies of the two-point discrete space give an infinite discrete space, not compact.

Where this goes next

The fundamental group of the circle uses the Lebesgue number lemma on I and on I \times I to lift paths and homotopies one small piece at a time, and Homotopy cites it wherever a square must be cut into small squares. Brouwer's fixed point theorem works on the closed disk D^2, compact by Theorem 4. In Classification of surfaces a surface glued from a polygon is compact, the polygon being compact by Theorem 4.

Exercises

Closed and bounded

  1. Decide which sets are compact, with a reason. (a) \{(x, y) : x^2 - xy + y^2 \le 3\}; (b) \{(x, y, z) : x^2 + y^2 - z^2 = 1\}; (c) \{(x, y) : 0 < x^2 + y^2 \le 1\}; (d) \{(x, y) : e^x + e^y \le 2\}.
  2. Find the points of the set in Exercise 1(a) farthest from the origin, and their distance.
  3. With the matrix with rows (a, b) and (c, d) as the point (a, b, c, d) \in \mathbb{R}^4, decide which sets are compact. (a) \det A = 1; (b) A^2 = A; (c) every entry in [0, 1] and \det A = 0. On the compact one, find the largest value of the trace a + d and where it occurs.
  4. Let N = \{(x, y) : |x|(1 + y^2) < 1\}. (a) Find the largest r with (-r, r) \times [-2, 3] \subseteq N. (b) Show that N is open and contains \{0\} \times \mathbb{R} but no tube W \times \mathbb{R}.
  5. Which are compact? (a) [0, 1] \times (0, 1]; (b) S^1 \times [-1, 1] \subseteq \mathbb{R}^3; (c) ([0, 1] \cap \mathbb{Q}) \times [0, 1]; (d) K \times [0, 1] with \displaystyle K = \{0\} \cup \{\frac{1}{n} : n \ge 1\}.

Lebesgue numbers and distances

  1. Naomi’s second bench is [0, 3] \times [0, 1], under zones U_i \times V_j with U_1 = (-1, 1.4), U_2 = (1, 2.2), U_3 = (1.9, 4), V_1 = (-1, 0.7), V_2 = (0.45, 2). Find the largest Lebesgue number of this cover, and two points that show nothing larger works.
  2. (a) Find the largest Lebesgue number of the cover of [0, 1] by \displaystyle [0, \frac{3}{5}) and \displaystyle (\frac{2}{5}, 1]. (b) Show that the sets \displaystyle (\frac{1}{n + 2}, \frac{1}{n}), n \ge 1, form an open cover of (0, 1) with no Lebesgue number.
  3. Let f(s) = (\cos 2\pi s, \sin 2\pi s) on I, \displaystyle U = \{(x, y) \in S^1 : y > -\frac{1}{2}\} and \displaystyle V = \{(x, y) \in S^1 : y < \frac{1}{2}\}. (a) Find f^{-1}(U) and f^{-1}(V). (b) Find the largest Lebesgue number of \{f^{-1}(U), f^{-1}(V)\}. (c) For which n from 1 to 7 does each piece \displaystyle [\frac{k - 1}{n}, \frac{k}{n}] lie in f^{-1}(U) or in f^{-1}(V)?
  4. Find the distance between the circle x^2 + (y - 4)^2 = 1 and the parabola y = x^2, and explain why Example 4 promises it is positive.

Proofs

  1. Prove that if X \times Y is compact and Y \ne \varnothing, then X is compact. Show that Y \ne \varnothing is needed.
  2. Prove that if Y is compact, \mathrm{pr}_1 : X \times Y \to X is a closed map. Show that \mathrm{pr}_1 : \mathbb{R} \times \mathbb{R} \to \mathbb{R} is not.
  3. Prove that if K \subseteq \mathbb{R}^n is compact and C \subseteq \mathbb{R}^n is closed, then K + C = \{k + c : k \in K, c \in C\} is closed. Find two closed subsets of \mathbb{R}^2 whose sum is not closed.
  4. Let f : \mathbb{R}^n \to \mathbb{R} be continuous with f(x) \to \infty as \lVert x \rVert \to \infty: for every M there is R with f(x) > M whenever \lVert x \rVert > R. Prove that f has a least value.
  5. Let d'(x, y) = \min\{\lVert x - y \rVert, 1\} on \mathbb{R}^n. Prove that d' is a metric with the same open sets as the Euclidean metric, and that \mathbb{R}^n is closed and bounded for d' but not compact.
Answers
  1. (a) Compact: closed, and \displaystyle x^2 - xy + y^2 - \frac{1}{2}(x^2 + y^2) = \frac{1}{2}(x - y)^2 \ge 0 gives x^2 + y^2 \le 6. (b) Not compact: unbounded, holding \displaystyle (\sqrt{1 + t^2}, 0, t) for every t. (c) Not compact: \displaystyle (\frac{1}{n}, 0) \to (0, 0), outside, so not closed. (d) Not compact: unbounded, since (-m, 0) has e^{-m} + 1 \le 2.
  2. Distance \displaystyle \sqrt{6}, at \displaystyle (\sqrt{3}, \sqrt{3}) and \displaystyle (-\sqrt{3}, -\sqrt{3}). By 1(a), x^2 + y^2 \le 6, with equality only if x = y and x^2 - xy + y^2 = 3, that is x^2 = 3.
  3. (a) Not compact: \displaystyle \operatorname{diag}(t, \frac{1}{t}) has determinant 1 and norm above t. (b) Not compact: rows (1, t), (0, 0) square to themselves for every t. (c) Compact: closed and inside [0, 1]^4. Largest trace 2, only at the matrix of all ones: a + d = 2 forces a = d = 1, then bc = 1 forces b = c = 1.
  4. (a) \displaystyle r = \frac{1}{10}: the largest 1 + y^2 on [-2, 3] is 10. (b) N is the preimage of (-1, 1) under a continuous function, and x = 0 gives 0. If (-r, r) \subseteq W, the point \displaystyle (\frac{r}{2}, \sqrt{2/r}) gives \displaystyle \frac{r}{2}(1 + \frac{2}{r}) = 1 + \frac{r}{2} > 1.
  5. (a) No: \displaystyle (\frac{1}{2}, \frac{1}{n}) \to (\frac{1}{2}, 0), outside, so not closed. (b) Yes: S^1 and [-1, 1] are compact (Theorem 4), so the product is (Theorem 2). (c) No: by Exercise 10 the factor [0, 1] \cap \mathbb{Q} would be compact, but \displaystyle \frac{\sqrt{2}}{2} is a limit point outside it. (d) Yes: K is closed and bounded, so compact, and Theorem 2 applies.
  6. \displaystyle \frac{1}{4}. As in Example 2, the overlaps are 0.4 and 0.3 across and 0.25 up. The points (0, 0.45) and (0, 0.7) are \displaystyle \frac{1}{4} apart, and only V_1 contains 0.45, which misses 0.7.
  7. (a) \displaystyle \frac{1}{5}. A set in neither member holds a point \displaystyle \ge \frac{3}{5} and one \displaystyle \le \frac{2}{5}; \displaystyle \{\frac{2}{5}, \frac{3}{5}\} shows nothing larger works. (b) Each point of \displaystyle [\frac{1}{n + 1}, \frac{1}{n}) lies in the n-th set, and \displaystyle \frac{1}{n + 1} in that set only. So \displaystyle \{\frac{1}{n + 2}, \frac{1}{n + 1}\} lies in no member, and its diameter \displaystyle \frac{1}{(n + 1)(n + 2)} tends to 0.
  8. (a) \displaystyle f^{-1}(U) = I \setminus [\frac{7}{12}, \frac{11}{12}] and \displaystyle f^{-1}(V) = I \setminus [\frac{1}{12}, \frac{5}{12}], from \displaystyle \sin 2\pi s \le -\frac{1}{2} and \displaystyle \sin 2\pi s \ge \frac{1}{2}. (b) \displaystyle \frac{1}{6}, the gap from \displaystyle \frac{5}{12} to \displaystyle \frac{7}{12}. (c) n = 2, 4, 6, 7 work, and n = 1, 3, 5 fail: [0, 1], \displaystyle [\frac{1}{3}, \frac{2}{3}] and \displaystyle [\frac{2}{5}, \frac{3}{5}] hold \displaystyle \frac{5}{12} and \displaystyle \frac{7}{12}.
  9. \displaystyle \frac{\sqrt{15}}{2} - 1. For p = (x, x^2) the squared distance to (0, 4) is u + (u - 4)^2 with u = x^2, least at \displaystyle u = \frac{7}{2}, value \displaystyle \frac{15}{4}; since \displaystyle \frac{\sqrt{15}}{2} > 1, the distance from p to the circle is \lVert p - (0, 4) \rVert - 1. Disjoint, as every parabola point is more than 1 from the center; the circle is compact and the parabola closed.
  10. Proof. \mathrm{pr}_1 is continuous (Theorem 4(c) of Subspace and product topologies) and onto X: for x \in X pick any y \in Y. So X = \mathrm{pr}_1(X \times Y) is compact (Theorem 5 of Compactness and open covers). If Y = \varnothing, then \mathbb{R} \times Y = \varnothing is compact while \mathbb{R} is not. \square
  11. Proof. Let C be closed and x_0 \notin \mathrm{pr}_1(C). Then \{x_0\} \times Y lies in the open set (X \times Y) \setminus C, so by Theorem 1 some open W \ni x_0 has W \times Y disjoint from C, and W misses \mathrm{pr}_1(C). So X \setminus \mathrm{pr}_1(C) is open. The closed hyperbola xy = 1 projects onto \mathbb{R} \setminus \{0\}, which is not closed. \square
  12. Proof. Let k_m + c_m \to z. A subsequence k_{m_j} \to k \in K (Theorem 8 of Compactness and open covers). Then c_{m_j} \to z - k, and z - k \in C because C is closed (Theorem 5(a) of Metric spaces, a first look). So z \in K + C, and K + C is closed by the same theorem. \square Example: A = \{(x, y) : x \ge 0, xy \ge 1\} and the x-axis. Their sum is \{(u, v) : v > 0\}, since \displaystyle (u, v) = (\frac{1}{v}, v) + (u - \frac{1}{v}, 0) and A has y > 0; it is not closed.
  13. Proof. Take M = f(0) and its R. The set S = \{x : f(x) \le f(0)\} is closed, contains 0, and lies in the closed ball of radius R, so it is compact (Theorem 4) and f has a least value on it at some c (Theorem 7 of Compactness and open covers). Points outside S have f(x) > f(0) \ge f(c). \square
  14. Proof. Symmetry and d'(x, y) = 0 \iff x = y are clear. If d'(x, y) or d'(y, z) is 1, then d'(x, y) + d'(y, z) \ge 1 \ge d'(x, z); otherwise the sum is \lVert x - y \rVert + \lVert y - z \rVert \ge d'(x, z). For r \le 1 the d'-ball of radius r is the Euclidean one, and a set is open for either metric exactly when each of its points has a ball of some radius r \le 1 inside it, so the open sets agree. \mathbb{R}^n is closed in itself and lies in the d'-ball of radius 2 about 0. It is not compact: compactness depends only on the open sets, and the balls B(0, m) cover with no finite subcover. \square
math.wiki · updated 2026-11-18 · CC-BY-SA-4.0