NiYaAGT.dvi @ Applied General Topology c© Universidad Politécnica de Valencia Volume 8, No. 2, 2007 pp. 151-159 Products of straight spaces with compact spaces Kusuo Nishijima and Kohzo Yamada Dedicated to the memory of Jan Pelant Abstract. A metric space X is called straight if any continuous real-valued function which is uniformly continuous on each set of a finite cover of X by closed sets, is itself uniformly continuous. Let C be the convergent sequence {1/n : n ∈ N} with its limit 0 in the real line with the usual metric. In this paper, we show that for a straight space X, X × C is straight if and only if X × K is straight for any compact metric space K. Furthermore, we show that for a straight space X, if X × C is straight, then X is precompact. Note that the notion of straightness depends on the metric on X. Indeed, since the real line R with the usual metric is not precompact, R × C is not straight. On the other hand, we show that the product space of an open interval and C is straight. 2000 AMS Classification: Primary 54C30, 54D05; Secondary 54C20, 54E35 Keywords: Straight spaces, convergent sequence, precompact, uniformly continuous. 1. Introduction All spaces are metric spaces and one fixed metric on a space X will be denoted by dX , and C(X) denotes the set of all continuous real-valued functions of a space X. Let (X, dX ) and (Y, dY ) be metric spaces. For a subspace M of X, we consider the restriction dX|M×M to M × M as a metric on M , which is denoted by dM . A metric dX×Y on the product space X × Y will be defined by dX×Y ((x1, y1), (x2, y2)) = √ (dX (x1, x2))2 + (dY (y1, y2))2. In this paper we study notions that use metrics in their definitions. However, the symbols of metrics will simply be denoted by d or be often omitted except when it is necessary to be clear which metric we consider. Let X be a metric space and {Fi : i = 1, 2, . . . , n} be a finite closed cover of X. Then it is well-known that every function f on X is continuous if the 152 K. Nishijima and K. Yamada restriction f|Fi of f is continuous on Fi for each i = 1, 2, . . . , n. However, it is not valid for uniform continuity. Indeed, consider the subspace X = {eiθ : 0 < θ < 2π} of the complex plane with the Euclidean metric and the function f (eiθ) = θ defined on X. Then the function f is not uniformly con- tinuous on X, but its restrictions on {eiθ : 0 < θ ≤ π} and {eiθ : π ≤ θ < 2π} are uniformly continuous. The following facts are useful to determine whether a given continuous function on a metric space is uniformly continuous or not. Lemma 1.1. Let f ∈ C(X). Then the following are equivalent: (1) f is uniformly continuous; (2) for every pair of sequences {xn} and {yn} in X if lim n→∞ d(xn, yn) = 0, then lim n→∞ |f (xn) − f (yn)| = 0. (3) for every pair of sequences {xn} and {yn} in X if lim n→∞ d(xn, yn) = 0, then there are subsequences {xkn } of {xn} and {ykn } of {yn} such that lim n→∞ |f (xkn ) − f (ykn )| = 0. Applying Lemma 1.1, it is easy to see that the above function f (eiθ) = θ is not u.c., because let αn = π n and βn = 2π − π n for each n ∈ N, and if we consider the sequences {xn = eiαn } and {yn = eiβn }, then lim n→∞ d(xn, yn) = 0, but lim n→∞ |f (xn) − f (yn)| = 2π. Recently, Berarducci, Dikranjan and Pelant [3] defined the following notion. Definition 1.2 ([3]). A metric space X is straight if whenever X is the union of finitely many closed sets, then f ∈ C(X) is uniformly continuous (briefly, u.c.) iff its restriction to each of the closed sets is u.c. Recall that a metric space X is called U C [1, 2] provided every continuous function on X is u.c. and a metric space is called uniformly locally connected if for every ε > 0 there is δ > 0 such that any two points at distance < δ lie in a connected set of diameter < ε. Clearly, all compact spaces are U C and all U C spaces are straight. Berarducci, Dikranjan and Pelant [3] prove that all uniformly locally connected spaces are straight. Hence, since the real line R and an open interval in R with the usual metric are clearly uniformly locally connected, they are straight, and of course, they are not U C. The product space of two compact spaces is compact, and hence U C. How- ever, in general, the product space X ×Y of a U C space X and a compact space Y need not be U C. Indeed, Atsuji’s result [2, Theorem 6] yields that if the product space X × Y of a non-compact and non-uniformly discrete U C space X and a space Y is U C, then Y must be uniformly discrete or finite (recall that a space is uniformly discrete if there is δ > 0 such that any two distinct points are at distance at least δ). On the other hand, there are non-compact and non-uniformly discrete straight spaces whose products with compact spaces are straight, for example, R × I is uniformly locally connected, and hence straight, where I means that the unit closed interval. Products of straight spaces with compact spaces 153 In this paper, we consider properties of a straight space whose product with any compact space is straight. Let X be a straight space and C be the convergent sequence {1/n : n ∈ N} with its limit 0 in the real line with the usual metric. Recall that a metric space X is precompact if for every ε > 0 there are finite points x1, x2, . . . , xn in X such that X = n ⋃ k=1 Bε(xk), where Bε(x) = {z ∈ X : d(x, z) < ε}. Then we will show the following: (1) X × C is straight if and only if X × K is straight for any compact space K; (2) if X × C is straight, then X is precompact. We can know, from the result (2), that R × C is not straight. On the other hand, we prove that the product space of an open interval and C is straight. However, we cannot decide whether the inverse implication of the result (2) is valid or not (cf. Acknowledgement). 2. Results We first introduce the terminology that is defined in [3]. Let X be a metric space. A pair E and F of closed sets of X is u-placed if d(Eε, Fε) > 0 for every ε > 0, where Eε = {x ∈ E : d(x, E ∩ F ) ≥ ε} and Fε = {x ∈ F : d(x, E ∩ F ) ≥ ε}. Note that if E ∩ F = ∅, then Eε = E and Fε = F . Hence, a partition X = E ∪ F of X into clopen sets is u-placed iff d(E, F ) > 0. Berarducci and Dikranjan and Pelant give the following characterizations of straight spaces in the same paper. Theorem 2.1 ([3]). For a metric space X the following are equivalent: (1) X is straight; (2) whenever X is the union of two closed sets, then f ∈ C(X) is u.c. iff its restriction to each of the closed sets is u.c.; (3) every pair of closed subsets, which form a cover of X, is u-placed. According to Theorem 2.1, we can conclude that the space Q of rational numbers and the space R \ Q of irrational numbers with the usual metric are not straight. Applying Lemma 1.1 and Theorem 2.1, we will show the following, which says that for given straight space X, it suffices to check whether X × C is straight in order to know whether X × K is straight for any compact space K. Theorem 2.2. For a straight space X X × C is straight if and only if X × K is straight for any compact space K. Proof. Assume that X × C is straight and let K be a compact space. From the definition of the straightness we assume that K is an infinite compact space. To show that X × K is straight, take a closed cover {E, F } of X × K and f ∈ C(X × K) on X × K such that the restrictions f|E and f|F are u.c. If we can show that f is u.c., then, from Theorem 2.1, our proof is complete. 154 K. Nishijima and K. Yamada Consider sequences {xn} and {yn} in X ×K such that lim n→∞ dX×K (xn, yn) = 0. We shall find subsequences {xkn } of {xn} and {ykn } of {yn} such that lim n→∞ |f (xkn ) − f (ykn )| = 0. We denote the projection of X × K onto K by πK . We consider the following cases. Case 1: πK ({xn : n ∈ ω}) is a finite set. Take a subsequence {xkn } of {xn} and z ∈ K such that πK (xkn ) = z for each n ∈ ω. Case 1.1: πK ({ykn : n ∈ ω}) is a finite set. In this case, since d(xkn , ykn ) converges to 0, there is an infinite subset Y ⊆ {ykn : n ∈ ω} for which πK (Y ) = {z}. So we may assume that {ykn : n ∈ ω} is the infinite set. Put Ez = E∩(X×{z}) and Fz = F ∩(X×{z}). Then we can know that (i) {Ez, Fz} is a closed cover of X × {z} and (ii) the restrictions f|Ez and f|Fz are u.c. Since X is straight and isometric to X ×{z}, X ×{z} is straight. It follows that the restriction f|X×{z} is u.c. Observe that the sequences {xkn } and {ykn } lie in X × {z} and lim n→∞ dX×{z}(xkn , ykn ) = 0. Hence, we have that lim n→∞ |f (xkn ) − f (ykn )| = lim n→∞ |f|X×{z}(xkn ) − f|X×{z}(ykn )| = 0. Case 1.2: πK ({ykn : n ∈ ω}) is an infinite set. Since K is compact, πK ({ykn : n ∈ ω}) contains a non-trivial convergent sequence. We may assume that πK ({ykn : n ∈ ω}) is the non-trivial conver- gent sequence and also πK (ykm ) 6= πK (ykn ) if m 6= n. Note that d(xkn , ykn ) converges to 0. Hence, it follows that z is the convergent point of the sequence {πK (ykn )}. Put zn = πK (ykn ) for each n ∈ ω and Z = {zn : n ∈ ω} ∪ {z}. Define a mapping g : X × C → X × Z by g(x, 1/n) = (x, zn) for each x ∈ X and n ∈ ω, and g(x, 0) = (x, z) for each x ∈ X. Clearly, g is a uniformly homeomorphism. Put H = g−1(E ∩ (X × Z)), I = g−1(F ∩ (X × Z)), akn = g −1(xkn ), bkn = g −1(ykn ) for each n ∈ ω, and h = f ◦ g : X × C → R. Then we can show the following: (i) {H, I} is a closed cover of X × C, (ii) lim n→∞ dX×C (akn , bkn ) = 0, and (iii) h|H = f|E∩(X×Z) ◦ g|H , and h|I = f|F ∩(X×Z) ◦ g|I , and hence h|H and h|I are u.c. Products of straight spaces with compact spaces 155 Since X × C is straight, h is u.c. Hence lim n→∞ |h(akn ) − h(bkn )| = 0. It follows that lim n→∞ |f (xkn ) − f (ykn )| = lim n→∞ |f (g(akn )) − f (g(bkn ))| = lim n→∞ |h(akn ) − h(bkn )| = 0. Case 2: πK ({xn : n ∈ ω}) is an infinite set. Since X is compact, we can pick a subsequence {xkn } of {xn} and z ∈ K such that πK (xkm ) 6= πK (xkn ) if m 6= n and {πK (xkn )} converges to z. Note that lim n→∞ d(xkn , ykn ) = 0. Hence, this yields that the sequence {πK (ykn )} also converges to z. Let {zn : n ∈ ω} be an enumeration of πK ({xkn : n ∈ ω}∪{ykn : n ∈ ω}) such that zm 6= zn if m 6= n. Then, the sequence {zn} converges to z. Consider the same mapping g : X × C → X × ({zn : n ∈ ω} ∪ {z}) as in Case 1.2. Then, with the same argument in Case 1.2, if we put akn = g −1(xkn ) and bkn = g −1(ykn ) for each n ∈ ω and h = g ◦ f , then we can show that h is u.c., and hence lim n→∞ |h(akn ) − h(bkn )| = 0. Consequently, lim n→∞ |f (xkn ) − f (ykn )| = lim n→∞ |f (g(akn )) − f (g(bkn ))| = lim n→∞ |h(akn ) − h(bkn )| = 0. Therefore, in any case, we can find subsequences {xkn } of {xn} and {ykn } of {yn} such that lim n→∞ |f (xkn ) − f (ykn )| = 0. It follows, from Lemma 1.1, that f is u.c. Consequently, X × K is straight. � The following result gives a necessary condition of X for which X × C is straight. Theorem 2.3. For a straight space X if X × C is straight, then X is precom- pact. Proof. Put Y = X × C. Suppose that X is not precompact and pick ε > 0 and an infinite set {xn : n ∈ N} such that Bε(xm) ∩ Bε(xn) = ∅ if m 6= n. For each n ∈ N let an = (xn, 1 n ) ∈ Y and bn = (xn, 1 n + 1 ) ∈ Y . Clearly, lim n→∞ dY (an, bn) = 0. Hence, we can find N ∈ N such that bn ∈ Bε/2(an) for every n ≥ N . Put M = Y \ ⋃ n≥N Bε(an). Then M is a closed subset of Y . For each n ≥ N put An = (X × { 1 i : i ≤ n}) ∩ Bε(an), Bn = (X × ({ 1 i : i ≥ n + 1} ∪ {0})) ∩ Bε(an). 156 K. Nishijima and K. Yamada Note that the collection {Bε(an) : n ∈ N} is closed discrete in Y , and hence so are {An : n ∈ N} and {Bn : n ∈ N}. If we put E = M ∪ ⋃ n≥N An and F = M ∪ ⋃ n≥N Bn, then we have that (a) {E, F } is a closed cover of Y , (b) E ∩ F = M , (c) for each n ≥ N d(an, E ∩ F ) = d(an, M ) ≥ d(an, Y \ Bε(an)) = ε, and (d) for each n ≥ N d(bn, E ∩ F ) = d(bn, M ) ≥ d(bn, Y \ Bε(an)) = ε 2 , because bn ∈ Bε/2(an). The conditions (c) and (d) imply that {an : n ≥ N} ⊆ Eε/2 and {bn : n ≥ N} ⊆ Fε/2. Since lim n→∞ d(an, bn) = 0, it follows that d(Eε/2, Fε/2) = 0. That is, the pair E and F is not u-placed. Consequently, by Theorem 2.1 we can prove that X × C is not straight. � Remark 2.4. Since R is a straight space that is not precompact, Theorem 2.3 says that R × C is not straight. Indeed, we can construct a pair of closed sets E and F which is not u-placed. For example, let E = ⋃ n∈N ([2n, ∞) ×{ 1 n }) and F = ⋃ n∈N ((−∞, 2n]×{ 1 n })∪R×{0}. Then {E, F } is a closed cover of R×C and E ∩ F = {(2n, 1 n ) : n ∈ N}. Put xn = (2n + 1, 1 n ) and yn = (2n + 1, 1 n + 1 ) for each n ∈ N. Then we can see that lim n→∞ d(xn, yn) = 0, {xn : n ∈ N} ⊆ E1/2 and {yn : n ∈ N} ⊆ F1/2. Hence d(E1/2, F1/2) = 0. This means that the pair E and F is not u-placed. Corollary 2.5. For a complete straight space X the following are equivalent: (1) X is precompact; (2) X is compact; (3) X × C is straight; (4) X × K is straight for any compact space K. We don’t know whether the inverse implication of Theorem 2.3 is true or not, however, we can show that the product space of an open interval and C is straight (cf. Acknowledgment). We need the following lemmas. Lemma 2.6. A metric space X which is represented as a topological sum of a family {Xα : α ∈ A} of a spaces is straight if inf{d(Xα, Xβ ) : α 6= β} > 0. The following lemma is introduced in [3, Theorem 5.3] and proved in [4, Proposition 2.4]. Lemma 2.7 ([3, 4]). Let X be a metric space and X = K ∪ Y , where K is a compact subspace of X and Y is a closed subset of X. Then X is straight iff Y is straight. Theorem 2.8. The product space of a half open interval and C is straight. Products of straight spaces with compact spaces 157 Proof. Let X = (a, b] × C, where a < b. To show that X is straight, let E and F be closed sets in X with E ∪ F = X and take an arbitrary (small) positive number ε > 0. To avoid confusion we use notations such as EXε and (E ∩ Y )Yε , and which mean that EXε = {x ∈ E : dX (x, E ∩ F ) ≥ ε} and (E ∩ Y )Yε = {x ∈ E ∩ Y : dY (x, E ∩ F ∩ Y ) ≥ ε}, where E, F and Y are subsets of a space X. According to Theorem 2.1, we shall show that the pair E and F is u-placed. Assuming that b = a + 1 and we can pick N ∈ N for which 1 N + 1 < ε√ 2 ≤ 1 N , put U = (a, a + ε√ 2 ) × ({ 1 n : n ≥ N + 1} ∪ {0}) and Y = X \ U. Case 1. U ∩ (E ∩ F ) 6= ∅. In this case, since the diameter of U is less than ε, U ⊆ BXε (E ∩ F ∩ U ) ⊆ BXε (E ∩ F ). Thus (2.1) EXε ∪ F Xε ⊆ X \ U = Y. It follows that dX (E X ε , F X ε ) = dY (E X ε , F X ε ). To show that E X ε ⊆ (E ∩ Y )Yε , let x ∈ EXε . Then x ∈ E and dX (x, E ∩ F ) ≥ ε. Since x ∈ E ∩ Y by (2.1) and dY (x, (E ∩ Y ) ∩ (F ∩ Y )) ≥ dX (x, E ∩ F ) ≥ ε, we can see that x ∈ (E ∩ Y )Yε . Therefore EXε ⊆ (E ∩ Y )Yε . In the same way, we can show that F Xε ⊆ (F ∩ Y )Yε . On the other hand, Lemma 2.6 and Lemma 2.7 yield that Y is straight, and hence dY ((E ∩ Y )Yε , (F ∩ Y )Yε ) > 0. So, we can get that dX (E X ε , F X ε ) ≥ dX ((E ∩ Y )Yε , (F ∩ Y )Yε ) = dY ((E ∩ Y )Yε , (F ∩ Y )Yε ) > 0. Case 2. U ∩ (E ∩ F ) = ∅. In this case, for every p ∈ { 1 n : n ≥ N + 1} ∪ {0} (a, a + ε√ 2 ) × {p} ⊆ E ∪ F(2.2) (E ∩ F ) ∩ ((a, a + ε√ 2 ) × {p}) = ∅.(2.3) Since every (a, a + ε√ 2 ) × {p} is connected, (a, a + ε√ 2 ) × {p} ⊆ E or (a, a + ε√ 2 ) × {p} ⊆ F. 158 K. Nishijima and K. Yamada Now, we assume that (a, a + ε√ 2 ) × {0} ⊆ E. Then, from the conditions (2.2) and (2.3), we can find M ≥ N + 1 such that (a, a + ε√ 2 ) × ({ 1 n : n ≥ M} ∪ {0}) ⊆ E. Put V = (a, a + ε 2 √ 2 ) × ({ 1 n : n ≥ M + 1} ∪ {0}) and Z = X \ V . Then E ∩ F ⊆ F ⊆ Z. Lemma 2.6 and Lemma 2.7 claim that Z is straight. So, we can say that (2.4) dZ ((E ∩ Z)Zε , (F ∩ Z)Zε ) > 0. Here, we shall show that (2.5) EXε ∩ Z ⊆ (E ∩ Z)Zε and F Xε ⊆ (F ∩ Z)Zε . Let x ∈ EXε ∩ Z. Then x ∈ E ∩ Z and dX (x, E ∩ F ) ≥ ε. Since E ∩ F ⊆ Z, dZ (x, (E ∩ Z) ∩ (F ∩ Z)) = dZ (x, E ∩ F ) = dX (x, E ∩ F ) ≥ ε. It follows that x ∈ (E ∩ F )Zε , and hence EXε ∩ Z ⊆ (E ∩ Z)Zε . Next, let x ∈ F Xε . Then x ∈ F and dX (x, E ∩ F ) ≥ ε. Since F ⊆ Z, x ∈ F ∩ Z and dZ (x, (E ∩ Z) ∩ (F ∩ Z)) = dZ (x, E ∩ F ) = dX (x, E ∩ F ) ≥ ε. It follows that x ∈ (F ∩ Z)Zε , and hence F Xε ⊆ (F ∩ Z)Zε . The conditions (2.4) and (2.5) yield that (2.6) dX (E X ε ∩ Z, F Xε ) ≥ dX ((E ∩ Z)Zε , (F ∩ Z)Zε ) = dZ ((E ∩ Z)Zε , (F ∩ Z)Zε ) > 0. Furthermore, since V = (a, a + ε 2 √ 2 ) × ({ 1 n : n ≥ M + 1} ∪ {0}) and ( (a, a + ε√ 2 ) × ( 1 n : n ≥ M} ∪ {0}) ) ∩ F = ∅, we can see that dX (V, F ) > 0, and hence (2.7) dX (E X ε ∩ V, F Xε ) > 0. The fact EXε = (E X ε ∩ V ) ∪ (EXε ∩ Z) and the conditions (2.6) and (2.7) yield that dX (E X ε , F X ε ) > 0. In any case, we can get dX (E X ε , F X ε ) > 0. Consequently, we can conclude that X = (a, b] × C is straight. With the same argument we can prove that [a, b) × C is also straight. � Corollary 2.9. The product space of an open interval and C is straight. Proof. Let X = (a, b) × C, where a < b. Take real numbers c and d for which a < c < d < b and put Y = (a, c] × C, Z = [d, b) × C and K = [c, d] × C. Then Theorem 2.7 and Lemma 2.6 yield that Y ∪Z is straight. Therefore, since Y ∪ Z is a straight closed subspace of X, K is compact and X = (Y ∪ Z) ∪ K, applying Lemma 2.8, we can show that X is straight. � Products of straight spaces with compact spaces 159 Finally, we obtain the following from Corollary 2.5 and Corollary 2.9. Corollary 2.10. The product of an open interval and a compact metric space is straight. Acknowledgements. When the second author gave a talk about the results in this paper at the III Japan-Mexico Joint Meeting in Topology and its Applications, he has learned from Professor Dikranjan that A. Berarducci, D. Dikranjan and J. Pelant in [4] proved independently that R × C is not straight and the product space of an open interval and C is straight. After that, Professor Dikranjan sent us their paper [5] quite recently, in which they obtain the interesting result such as the inverse implication of Theorem 2.3 is valid. Therefore, we can know now that for a straight space X X × K is straight for every compact metric space K if and only if X is precompact. References [1] M. Atsuji, Uniform continuity of continuous functions on metric spaces, Pacific J. Math. 8 (1958), 11–16. [2] M. Atsuji, Uniform continuity of continuous functions on metric spaces, Canad. J. Math. 13 (1961), 657–663. [3] A. Berarducci, D. Dikranjan and J. Pelant, An additivity theorem for uniformly contin- uous functions, Topology Appl. 146-147 (2005), 339–352. [4] A. Berarducci, D. Dikranjan and J. Pelant, Local connectedness and extension of uni- formly continuous functions, preprint. [5] A. Berarducci, D. Dikranjan and J. Pelant, Products of straight spaces, preprint. Received August 2005 Accepted August 2006 Kusuo Nishijima (lkusuo@ezweb.ne.jp) Faculty of Education, Shizuoka University, Shizuoka, 422-8529 Japan. Kohzo Yamada (eckyama@ipc.shizuoka.ac.jp) Faculty of Education, Shizuoka University, Shizuoka, 422-8529 Japan.