Math 516 Professor Lieberman February 6, 2009 HOMEWORK #2 SOLUTIONS Chapter 8 13. Let T be the weakest topology containing C . Then B ⊂ T . In addition, each x ∈ X is in some B ∈ B. (Specifically, every x is in X ∈ B.) If B1 and B2 are two elements of B, then we can write n m B1 = ⋂ C1i , B2 = ⋂ C2j , i=1 j=1 where n and m are positive integers and C1i and C2j are elements of C. It follows that B1 ∩ B2 = C11 ∩ ⋅ ⋅ ⋅ ∩ C1n ∩ C21 ∩ ⋅ ⋅ ⋅ ∩ C2m , so B1 ∩ B2 ∈ B. Therefore B is the base for some topology, which must be T . 14. First, X is in T because X = X ∖ ∅ and ∅ ∈ T by hypothesis. Second, if (Oα )α∈A is a family of sets in T , then X ∖ ∪Oα = ∩(X ∖ Oα ). To give a complete proof, you have to consider two possibilities. Either X ∖ Oα is infinite for all α, in which case Oα is empty for all α, so the union is the empty set, which is in T or X ∖ Oα is finite for α, in which case the intersection is a subset of a finite set and hence finite, so O ∖ ∪Oα ∈ T . Finally, if O1 , . . . , On are in T , then X ∖ ∩Oj = ∪(X ∖ Oj ). This union is a union of finitely many finite sets, so it’s in T . To see that T is not first countable, fix x ∈ X and let Y = X ∖ {x}. Because X is uncountable, so is Y . Now let (Uj ) be a countable collection of neighborhoods of x. It follows that X ∖ ∪Uj is still uncountable, so there is a point y ∈ Y such that y ∉ Uj for all positive integers j. Hence, for any positive integer j, Uj is not a subset of the open set O = X ∖ {y}. Since x ∈ O, it follows that (Uj ) is not a base for the topology at x, so T is not first countable. 15. If x ∈ R, then x ∈ [x, x + 1). If x ∈ [a1 , b1 ) ∩ [a2 , b2 ) for some real numbers a1 < b1 and a2 < b2 , then b1 > x and b2 > x, so x ∈ [x, min{b1 , b2 }). Hence, by Proposition 5, B is the base for a topology. For each x ∈ R, we can take the collection of all intervals of the form [x, x + n1 ) for some positive integer n as a base at x. Hence the half-open interval topology is first countable. On the other hand, to examine the second countability axiom, we let x ∈ R and let O be an open set, and we introduce the open set V = [x, x + 1). If x < inf O, then x ∉ O, and if x > inf O, then there is a number y ∈ (x, inf O). Hence if x ≠ inf O, we cannot have x ∈ O ⊂ V . Hence a base for the half-open topology 1 2 must contain at least one open set Ox for each x ∈ R. (This is a set Ox such that x ∈ Ox ⊂ [x, x + 1).) Since R is uncountable, it follows that any base for this topology is uncountable. Finally, if x ∈ R and if O is an open set containing x, then there is a positive integer n such that [x, x + n1 ) ⊂ O. Since there is a rational number r in the interval (x, x + x1 ), it follows that r ∈ O. Therefore Q is dense. it follows from Proposition 7.6 that this topology is not metrizable. 24. (a) This is simple algebra: ∣f ∣ 1 + ∣f ∣ < = 1. 1 + ∣f ∣ 1 + ∣f ∣ (b) First, h is continuous because it’s the composition of the continuous function g∶ R → R given by g(s) = s/(1 + ∣s∣) and the continuous function f . Because h is continuous, the sets B and C are closed. Urysohn’s Lemma then gives a continuous real-valued function H1 on X such that H1 (B) = {1} and H1 (C) = {0} and H1 (x) ∈ [0, 1] for all x ∈ X. We then choose h1 (x) = (1 − 2H1 (x))/3. It follows that h1 is a real-valued function on X. In addition h1 (x) = (1 − 2(1))/3 = −1/3 for all x ∈ B, h1 (x) = (1 − 2(0))/3 = 1/3 for all x ∈ C. Furthermore, if x ∈ X, then 1 1 1 h1 (x) = (1 − 2H1 (x)) ≥ (1 − 2(0)) = − 3 3 3 and 1 1 1 h1 (x) = (1 − 2H1 (x)) ≤ (1 − 2(1)) = , 3 3 3 so ∣h1 (x)∣ ≤ 1/3. Finally, if x ∈ A, we consider three possibilities. If also x ∈ B, then h(x) − h1 (x) ≤ − 31 − (− 13 ) = 0 ≤ 23 and h(x) − h1 (x) ≥ −1 − (− 13 ) = − 23 . If also x ∈ C, then h(x) − h1 (x) ≤ 1 − ( 31 ) = 23 and h(x) − h1 (x) ≥ 13 − ( 31 ) = 0 ≥ − 23 . If x ∈ A ∖ (B ∪ C), then ∣h(x) − h1 (x)∣ ≤ ∣h(x)∣ + ∣h1 (x)∣ < 31 + 31 = 23 . In all cases, we have ∣h(x) − h1 (x)∣ ≤ 23 . (c) We proceed by induction on n. Part (b) gives the case n = 1. Once we have h1 , . . . , hn , we construct hn+1 as follows: Let n n 2n 2n Bn = {x ∶ h(x) − ∑ hi (x) ≤ − n+1 }, Cn = {x ∶ h(x) − ∑ hi (x) ≥ n+1 }, 3 3 i=1 i=1 ∣h∣ = and define the function Kn by Kn (x) = n 3n (h(x) − hi (x)) . ∑ 2n i=1 Applying part (b) to the function Kn (with Bn in place of B and Cn in place of C), we see that there is a continuous real-valued function kn such that kn (x) = − 31 for all x ∈ Bn , kn (x) = 13 for all x ∈ Cn , ∣kn (x)∣ ≤ 13 for all x ∈ X, and ∣Kn (x)−kn (x)∣ < 23 for all x ∈ A. We now take hn+1 = (2n /3n )kn . Then ∣hn+1 (x)∣ ≤ (2n /3n ) 13 = 2n /3n+1 for all x ∈ X, and n+1 ∣h(x) − ∑ hi (x)∣ = i=1 2n 2n+1 ∣K (x) − k (x)∣ < . n n 3n 3n+1 3 (d) This part is just the Weierstrass M -test. (For completeness, I will include a proof here, but it isn’t really needed.) First, we show that the sequence of partial sums (sn ), given by n sn (x) = ∑ hn (x), k=1 is uniformly Cauchy. For ε > 0 be given, there is a positive integer N such that ∞ ∑n=N 2n−1 /3n < ε (because this is the tail of a geometric series with ratio 2/3). If n > m ≥ N , it follows that n n n i=m+1 i=m+1 i=m+1 ∣sn (x) − sm (x)∣ = ∣ ∑ hi (x)∣ ≤ ∑ ∣hi (x)∣ ≤ ∑ 2i−1 3i < ε. Hence, for each x ∈ X, there is a number k(x) such that sn (x) → k(x) because R is complete. Now, let ε > 0 be given and let N be the positive integer such that ∣sn (x) − sm (x)∣ < ε/2 for all n, m ≥ N and all x ∈ X. Fix an x ∈ X and note that there is a positive integer N (x) such that ∣sn (x) − k(x)∣ < ε/2 if n ≥ N (x). If n = max{N, N (x)} and if m ≥ N , it follows that ∣sm (x) − k(x)∣ ≤ ∣sm (x) − sn (x)∣ + ∣sn (x) − k(x)∣ < ε. Hence the sequence (sn ) converges uniformly to k. Finally, we show that k is continuous at any x ∈ X. Let ε > 0 and x ∈ X be given. Then there is a positive integer N such that ∣sn (t) − k(t)∣ < ε/3 for any t ∈ X and there is a neighborhood O of x such that ∣sN (x) − sN (y)∣ < ε/3 for all y ∈ O. If y ∈ O, it follows that ∣k(x) − k(y)∣ ≤ ∣k(x) − sN (x)∣ + ∣sN (x) − sN (y)∣ + ∣sN (y) − k(y)∣ < ε. Hence k is continuous at x. (e) Since k is continuous, it follows that {k(x) = 1} is closed. Hence Urysohn’s Lemma gives a continuous function ϕ on X which is 1 on A and 0 on {k(x) = 1}. (f) Since ϕ = 0 when k = 1, it follows that ϕ∣k∣ < 1 on X. Therefore, g = ϕk/(1−ϕ∣k∣) is continuous on X. In addition, on A, we have k f /(1 + ∣f ∣) f g= = = = f. 1 − ∣k∣ 1 − (∣f ∣/(1 + ∣f ∣)) 1 + ∣f ∣ − ∣f ∣ 25. Here is some notation that makes the proof easier to read. For ε > 0, y ∈ X, and F ∗ a finite subset of F , we write B(ε, y, F ∗ ) = {x ∶ ∣f (x) − f (y)∣ < ε for all f ∈ F ∗ }. To show that B, the collection of all such B(ε, y, F ∗ ) is a base for the weak topology, we must show that every one of these sets is in the weak topology and that, if U is in the weak topology and x ∈ U , then there is a choice of ε, y and F ∗ so that x ∈ B(ε, y, F ∗ ) ⊂ U . The first step is easy. Fix ε > 0, y ∈ X and F ∗ a finite subset of F . For each fi ∈ F ∗ , we set Bi = B(ε, y, {fi }), so B(ε, y, F ∗ ) = ⋂ Bi . For Oi = (f (y) − ε, f (y) + ε), we then have that Bi = fi−1 (Oi ). Hence Bi is in subbase denoted by C (on page 179) for the weak topology, and therefore B(ε, y, F ∗ ) is also in the weak topology. 4 For the second step, we show that B is a base for some topology T . According to Proposition 5, we must show that each x ∈ X is in some B(ε, y, F ∗ ) and that, if x ∈ X is in the intersection of B(ε1 , y1 , F1∗ ) and B(ε2 , y2 , F3∗ ), then there is a B(ε3 , y3 , F3∗ ) such that x ∈ B(ε3 , y3 , F3∗ ) ⊂ B(ε1 , y1 , F1∗ ) ∩ B(ε2 , y2 , F2∗ ). The first statement is proved by noting that, for any x ∈ X and any f ∈ F , we have x ∈ B(1, x, {f }). To prove the second statement, we choose y3 = x and F3∗ = F1∗ ∪ F2∗ . Then, for any ε3 > 0, we have x ∈ B(ε3 , y3 , F3∗ ). We now set η1 = min∗ {∣f (x) − f (y)∣}, f ∈F1 η2 = min∗ {∣f (x) − f (y)∣}. f ∈F2 Since x, y1 , and y2 are fixed and F1∗ and F2∗ are finite, η1 ∈ (0, ε1 ) and η2 ∈ (0, ε2 ). We now take ε3 = min{ε1 − η1 , ε2 − η2 }. If z ∈ B(ε3 , y3 , F3∗ ), then, for any f ∈ F1∗ , we have ∣f (z) − f (y1 )∣ ≤ ∣f (z) − f (x)∣ + ∣f (x) − f (y1 )∣ < ε3 + ∣f (x) − f (y1 )∣ ≤ ε3 + η1 ≤ ε1 , and for any f ∈ F2∗ , we have ∣f (z) − f (y2 )∣ ≤ ∣f (z) − f (x)∣ + ∣f (x) − f (y2 )∣ < ε3 + ∣f (x) − f (y2 )∣ ≤ ε3 + η2 ≤ ε2 . Hence B(ε3 , y3 , F3∗ ) ⊂ B(ε1 , y1 , F1∗ ) ∩ B(ε2 , y2 , F2∗ ). Now, suppose that the weak topology is Hausdorff. Then for any pair of distinct points x and y in X, there are disjoint open sets (in the weak topology) O1 and O2 such that x ∈ O1 and y ∈ O2 . Hence there are ε1 and ε2 , y1 and y2 , and F1∗ and F2∗ such that x ∈ B(ε1 , y1 , F1∗ ), y ∈ B(ε2 , y2 , F2∗ ), and B(ε1 , y1 , F1∗ ) ∩ B(ε1 , y1 , F1∗ ) = ∅. Suppose that, for every function f ∈ F , we had f (x) = f (y). It follows that ∣f (x) − f (y1 )∣ < ε1 for all f ∈ F1∗ and ∣g(y) − g(y2 )∣ < ε2 for all g ∈ F2∗ . But then, g(x) = g(y) for all g ∈ F2 so x ∈ B(ε2 , y2 , F2∗ ) which contradicts our assumption that B(ε1 , y1 , F1∗ ) ∩ B(ε1 , y1 , F1∗ ) = ∅. Therefore there is some function f ∈ F with f (x) ≠ f (y). Conversely, if, for every x ≠ y in X, there is a function f ∈ F such that f (x) ≠ f (y), let ε = 21 ∣f (x) − f (y)∣. Then B(ε, x, {f }) and B(ε, y, {f }) are open sets with x ∈ B(ε, x, {f }) and y ∈ B(ε, y, {f }). If z ∈ B(ε, x, {f }), then ∣f (x) − f (z)∣ < ε, so ∣f (z) − f (y)∣ ≥ ∣f (x) − f (y)∣ − ∣f (y) − f (z)∣ = 2ε − ∣f (y) − f (z)∣ > ε. Hence z ∉ B(ε, y, {f }), which means that B(ε, x, {f }) and B(ε, y, {f }) are disjoint. Therefore the weak topology is Hausdorff in this case. 32. Let A and B be non-empty open subsets of G such that G = A ∪ B. Then, for each α, we have Cα ⊂ A∪B. If A∩Cα = ∅ for some α, then Cα ⊂ B. Since G∩A is non-empty, it follows that Cβ ∩ A is non-empty for some β. Since Cα ∩ Cβ is non-empty, there is a point x ∈ Cα ∩ Cβ , so x ∈ B. It follows that Cβ ∩ A and Cβ ∩ B are non-empty open subsets of Cβ and hence, because Cβ is connected, A ∩ B is non-empty. Hence G is connected, too. 42. Let x = (xα ) and y = (yα ) be disjoint points in the direct product. Then there is an index α0 such that xα0 ≠ yα0 . Because Xα0 is Hausdorff, there are disjoint open subsets Uα0 and Vα0 of Xα0 with xα0 ∈ Uα0 and yα0 ∈ Vα0 . 5 We now set Uα = { Uα0 Xα if α = α0 , otherwise, Vα Vα = { 0 Xα if α = α0 , otherwise. It follows that U = ⨉ Uα and V = ⨉ Vα are disjoint open subsets of ⨉ Xα with x ∈ U α α and y ∈ V . Hence the product is Hausdorff. α 6 Chapter 9 4. Since the proof of this fact is so similar to the proof of problem 3, I will prove both results, using X to denote the compact Hausdorff space. If x ∈ X and F is a nonempty closed subset of X such that x ∉ F , then, for each y ∈ F , there are disjoint open sets Uy and Vy such that x ∈ Uy and y ∈ Vy . (This follows from the Hausdorff property.) Hence (Vy )y∈F is an open cover of F . From Proposition 12, F is compact, and hence this open cover has a finite subcover Vy1 , . . . , Vyn . We then set U = ∩ni=1 Uyi , V = ∪ni=1 Vyi . Because Vy1 , . . . , Vyn is an open cover of F , it follows that V is open and F ⊂ V . Because x ∈ Uy for all y ∈ F , it follows that x ∈ U ; in addition, U is open because it’s a finite intersection of open sets. Finally, if z ∈ U , then, for each i, z ∈ Uyi so z ∉ Vyi and therefore z ∉ V . Therefore U and V are disjoint, so X is regular. Now, if A and B are disjoint, non-empty closed subsets of X, then, for each x ∈ A, the previous step shows that there are disjoint open sets Ux and Vx such that x ∈ Ux and B ⊂ Vx . Since A is compact, we can take a finite subcover Ux1 , . . . , Uxm . If we set U = ∪m V = ∩m i=1 Uxi , i=1 Vxi , then U and V are disjoint open sets with A ⊂ U and B ⊂ V . Therefore X is normal. 8. (a) Fix x ∈ X and let O be a neighborhood of f (x). Then {x} is compact, so there is a positive integer M such that n ≥ M implies that fn ∈ N{x},O . In other words, fn (x) ⊂ O. Therefore, lim fn (x) = f (x). (b) Suppose fn → f in the compact open topology, and let C be a compact subset of X. We first show that f∣C is continuous. Fix a point x ∈ C and let ε > 0 be given. Write Ox for the open ball B(f (x), ε/2). Then there is a positive integer N1 such that ρ(f (x), fn (x)) < ε/2 by part (a). Let K= fN−11 (B(f (x), ε/4)). Note that K is closed because it’s the inverse image of a closed set. Moreover, K ∩ C is compact, by Proposition 12, because it’s a closed subset of a compact set. Since fn → f in the compact-open topology, there is a positive integer N2 such that fn (K) ⊂ Ox . Then U = fN−11 (O) ∩ C is an open subset of C and U ⊂ K, so fn (U ) ⊂ B(f (x), ε/2) for n ≥ N2 . Hence, if y ∈ U and n ≥ N2 , it follows that ρ(fn (y), f (x)) < ε/2, In addition, there is a positive integer N3 (y) such that ρ(fn (y), f (y)) < ε/2 if n ≥ N3 (y). Now, we fix y ∈ U and n = max{N2 , N3 (y)} to conclude that ρ(f (x), f (y)) ≤ ρ(f (x), fn (y)) + ρ(fn (y), f (y)) < ε. Hence f∣C is continuous at x. To show that fn → f uniformly on a compact set C, we first fix x ∈ C and ε > 0. Then we set O = B(f (x), ε/2), U = f −1 (O) ∩ C, and K = f −1 (B(f (x), ε/4)). As before U is open and K is compact with U ⊂ K. Hence there is a number N4 such that ρ(f (x), fn (y)) < ε/2 for n ≥ N3 and y ∈ K. Hence for y ∈ K and n ≥ N4 , we have ρ(fn (y), f (y)) ≤ ρ(fn (y), f (x)) + ρ(f (x), f (y)) < ε. 7 Therefore, for any x ∈ C, there are an open set U (x) and a positive integer M (x) such that x ∈ U (x) and ρ(fn (y), f (y)) < ε for all y ∈ U (x) and n ≥ M (x). Hence (U (x))x∈C is an open cover of C so it has a finite subcover: U (x1 ), . . . U (xm ).