Salvagt.dvi @ Applied General Topology c© Universidad Politécnica de Valencia Volume 5, No. 1, 2004 pp. 129- 136 Fuzzy quasi-metric spaces Valent́ın Gregori and Salvador Romaguera∗ Abstract. We generalize the notions of fuzzy metric by Kramosil and Michalek, and by George and Veeramani to the quasi-metric set- ting. We show that every quasi-metric induces a fuzzy quasi-metric and, conversely, every fuzzy quasi-metric space generates a quasi-metrizable topology. Other basic properties are discussed. 2000 AMS Classification: 54A40, 54E35, 54E15. Keywords: Fuzzy quasi-metric space; Quasi-metric; Quasi-uniformity; Bi- complete; Isometry. 1. Introduction In [9], Kramosil and Michalek introduced and studied an interesting notion of fuzzy metric space which is closely related to a class of probabilistic met- ric spaces, the so-called (generalized) Menger spaces. Later on, George and Veeramani started, in [3] (see also [5]), the study of a stronger form of metric fuzziness. In particular, it is well known that every metric induces a fuzzy metric in the sense of George and Veeramani, and, conversely, every fuzzy metric space in the sense of George and Veeramani (and also of Kramosil and Michalek) generates a metrizable topology ([4], [6], [9], [11], [13]). On the other hand, it is also well known that quasi-metric spaces constitute an efficient tool to discuss and solve several problems in topological algebra, approximation theory, theoretical computer science, etc. (see [10]). In this paper, we introduce two notions of fuzzy quasi-metric space that generalize the corresponding notions of fuzzy metric space by Kramosil and Michalek, and by George and Veeramani to the quasi-metric context. Several basic properties of these spaces are obtained. We show that every quasi-metric induces a fuzzy quasi-metric and, conversely, every fuzzy quasi-metric generates a quasi-metrizable topology. With the help of these results one can easily derive many properties of fuzzy quasi-metric spaces. ∗The authors acknowledge the support of Generalitat Valenciana, grant GRUPOS 03/027. 130 V. Gregori and S. Romaguera Our basic references for quasi-uniform and quasi-metric spaces are [2] and [10]. Let us recall that a quasi-pseudo-metric on a set X is a nonnegative real valued function d on X × X such that for all x, y, z ∈ X : (i) d(x, x) = 0; (ii) d(x, z) ≤ d(x, y) + d(y, z). Following the modern terminology (see Section 11 of [10]), by a quasi-metric on X we mean a quasi-pseudo-metric d on X that satisfies the following con- dition: d(x, y) = d(y, x) = 0 if and only if x = y. If the quasi-pseudo-metric d satisfies: d(x, y) = 0 if and only if x = y, then we say that d is a T1 quasi-metric on X. A quasi-(pseudo-)metric space is a pair (X, d) such that X is a (nonempty) set and d is a quasi-(pseudo-)metric on X. The notion of a T1 quasi-metric space is defined in the obvious manner. Each quasi-pseudo-metric d on X generates a topology τd on X which has as a base the family of open d-balls {Bd(x, r) : x ∈ X, r > 0}, where Bd(x, r) = {y ∈ X : d(x, y) < r} for all x ∈ X and r > 0. Observe that if d is a quasi-metric, then τd is a T0 topology, and if d is a T1 quasi-metric, then τd is a T1 topology. A topological space (X, τ) is said to be quasi-metrizable if there is a quasi- metric d on X such that τ = τd. In this case, we say that d is compatible with τ, and that τ is a quasi-metrizable topology. Given a quasi-(pseudo-)metric d on X, then the function d−1 defined on X × X by d−1(x, y) = d(y, x), is also a quasi-(pseudo-)metric on X, called the conjugate of d. Finally, the function ds defined on X × X by ds(x, y) = max{d(x, y), d−1(x, y)} is a (pseudo-)metric on X. 2. Definitions and basic results According to [13], a binary operation ∗ : [0, 1]×[0, 1] → [0, 1] is a continuous t-norm if ∗ satisfies the following conditions: (i) ∗ is associative and commu- tative; (ii) ∗ is continuous; (iii) a ∗ 1 = a for every a ∈ [0, 1]; (iv) a ∗ b ≤ c ∗ d whenever a ≤ c and b ≤ d, with a, b, c, d ∈ [0, 1]. Definition 2.1. A KM-fuzzy quasi-pseudo-metric on a set X is a pair (M, ∗) such that ∗ is a continuous t-norm and M is a fuzzy set in X × X × [0, +∞) such that for all x, y, z ∈ X : (KM1) M(x, y, 0) = 0; (KM2) M(x, x, t) = 1 for all t > 0; (KM3) M(x, z, t + s) ≥ M(x, y, t) ∗ M(y, z, s) for all t, s ≥ 0; (KM4) M(x, y, ) : [0, +∞) → [0, 1] is left continuous. Definition 2.2. A KM-fuzzy quasi-metric on X is a KM-fuzzy quasi-pseudo- metric (M, ∗) on X that satisfies the following condition: (KM2’) x = y if and only if M(x, y, t) = M(y, x, t) = 1 for all t > 0. If (M, ∗) is a KM-fuzzy quasi-pseudo-metric on X satisfying: (KM2”) x = y if and only if M(x, y, t) = 1 for all t > 0, Fuzzy quasi-metric spaces 131 we say that (M, ∗) is a T1 KM-fuzzy quasi-metric on X. Definition 2.3. A KM-fuzzy (pseudo-)metric on X is a KM-fuzzy quasi- (pseudo-)metric (M, ∗) on X such that for each x, y ∈ X : (KM5) M(x, y, t) = M(y, x, t) for all t > 0. Remark 2.4. It is clear that every KM-fuzzy metric is a T1 KM-fuzzy quasi- metric; every T1 KM-fuzzy quasi-metric is a KM-fuzzy quasi-metric, and every KM-fuzzy quasi-metric is a KM-fuzzy quasi-pseudo-metric. Definition 2.5. A KM-fuzzy quasi-(pseudo-)metric space is a triple (X, M, ∗) such that X is a (nonempty) set and (M, ∗) is a KM-fuzzy quasi-(pseudo- )metric on X. The notions of a T 1 KM-fuzzy quasi-metric space and of a KM-fuzzy (pseudo- )metric space are defined in the obvious manner. Note that the KM-fuzzy metric spaces are exactly the fuzzy metric spaces in the sense of Kramosil and Michalek. If (M, ∗) is a KM-fuzzy quasi-(pseudo-)metric on a set X, it is immediate to show that (M−1, ∗) is also a KM-fuzzy quasi-(pseudo-)metric on X, where M−1 is the fuzzy set in X × X × [0, +∞) defined by M−1(x, y, t) = M(y, x, t). Moreover, if we denote by Mi the fuzzy set in X × X × [0, +∞) given by Mi(x, y, t) = min{M(x, y, t), M−1(x, y, t)}, then (Mi, ∗) is, clearly, a KM-fuzzy (pseudo-)metric on X. Proposition 2.6. Let (X, M, ∗) be a KM-fuzzy quasi-pseudo-metric space. Then, for each x, y ∈ X the function M(x, y, ) is nondecreasing. Proof. Let x, y ∈ X and 0 ≤ t < s. Then M(x, y, s) ≥ M(x, x, s − t) ∗ M(x, y, t) = M(x, y, t). � Given a KM-fuzzy quasi-pseudo-metric space (X, M, ∗) we define the open ball BM (x, r, t), for x ∈ X, 0 < r < 1, and t > 0, as the set BM (x, r, t) := {y ∈ X : M(x, y, t) > 1 − r}. Obviously, x ∈ BM (x, r, t). By Proposition 2.6, it immediately follows that for each x ∈ X, 0 < r1 ≤ r2 < 1 and 0 < t1 ≤ t2, we have BM(x, r1, t1) ⊆ BM(x, r2, t2). Consequently, we may define a topology τM on X as τM := {A ⊆ X : for each x ∈ A there are r ∈ (0, 1), t > 0, with BM (x, r, t) ⊆ A}. Moreover, for each x ∈ X the collection of open balls {BM(x, 1/n, 1/n) : n = 2, 3, ...}, is a local base at x with respect to τM . It is clear, that if (X, M, ∗) is a KM-fuzzy quasi-metric (respectively, a T1 KM-fuzzy quasi-metric, a KM-fuzzy metric), then τM is a T0 (respectively, a T1, a Hausdorff) topology. The topology τM is called the topology generated by the KM-fuzzy quasi- pseudo-metric space (X, M, ∗). Similarly to the proof of Result 3.2 and Theorem 3.11 of [3], one can show the following results. 132 V. Gregori and S. Romaguera Proposition 2.7. Let (X, M, ∗) be a KM-fuzzy quasi-pseudo-metric space. Then, each open ball BM (x, r, t) is an open set for the topology τM . Proposition 2.8. A sequence (xn)n in a KM-fuzzy quasi-pseudo-metric space (X, M, ∗) converges to a point x ∈ X with respect to τM if and only if limn M(x, xn, t) = 1 for all t > 0. Definition 2.9. A GV-fuzzy quasi-pseudo-metric on a set X is a pair (M, ∗) such that ∗ is a continuous t-norm and M is a fuzzy set in X × X × (0, +∞) such that for all x, y, z ∈ X, t, s > 0 : (GV1) M(x, y, t) > 0; (GV2) M(x, x, t) = 1; (GV3) M(x, z, t + s) ≥ M(x, y, t) ∗ M(y, z, s); (GV4) M(x, y, ) : (0, +∞) → (0, 1] is continuous. Definition 2.10. A GV-fuzzy quasi-metric on X is a GV-fuzzy quasi-pseudo- metric (M, ∗) on X such that for all t > 0: (GV2’) x = y if and only if M(x, y, t) = M(y, x, t) = 1. If (M, ∗) is a GV-fuzzy quasi-pseudo-metric on X such that for all t > 0: (GV2”) x = y if and only if M(x, y, t) = 1, we say that (M, ∗) is a T1 KM-fuzzy quasi-metric on X. Definition 2.11. A GV-fuzzy (pseudo-)metric on X is a GV-fuzzy quasi- (pseudo-)metric (M, ∗) on X such that for all x, y ∈ X, t > 0 : (KM5) M(x, y, t) = M(y, x, t). Remark 2.12. It is clear that every GV-fuzzy metric is a T1 GV-fuzzy quasi- metric; every T1 GV-fuzzy quasi-metric is a GV-fuzzy quasi-metric, and every GV-fuzzy quasi-metric is a GV-fuzzy quasi-pseudo-metric. Definition 2.13. A GV-fuzzy quasi-(pseudo-)metric space is a triple (X, M, ∗) such that X is a (nonempty) set and (M, ∗) is a GV-fuzzy quasi-(pseudo-)metric on X. The notions of a T 1 GV-fuzzy quasi-metric space and of a GV-fuzzy metric space are defined in the obvious manner. Note that the GV-fuzzy metric spaces are exactly the fuzzy metric spaces in the sense of George and Veeramani. Remark 2.14. Note that if (M, ∗) is a GV-fuzzy quasi-(pseudo-)metric on X, then the fuzzy sets in X × X × (0, +∞), M−1 and Mi given by M−1(x, y, t) = M(y, x, t) and Mi(x, y, t) = min{M(x, y, t), M−1(x, y, t)}, are, as in the KM- case, a GV-fuzzy quasi-(pseudo-)metric and a GV-fuzzy (pseudo-)metric on X, respectively. Thus, condition (GV2’) above is equivalent to the following: M(x, x, t) = 1 for all x ∈ X and t > 0, and Mi(x, y, t) < 1 for all x 6= y and t > 0. Remark 2.15. Obviously, each GV-fuzzy quasi-(pseudo-)metric (M, ∗) can be considered as a KM-fuzzy quasi-(pseudo-)metric by defining M(x, y, 0) = 0 for Fuzzy quasi-metric spaces 133 all x, y ∈ X. Therefore, each GV-fuzzy quasi-pseudo-metric space generates a topology τM defined as in the KM-case, and Propositions 2.6, 2.7 and 2.8 above remain valid for GV-fuzzy quasi-pseudo-metric spaces. Example 2.16 (compare Example 2.9 of [3]). áLet (X, d) be a quasi-metric space. Denote by a · b the usual multiplication for every a, b ∈ [0, 1], and let Md be the function defined on X × X × (0, +∞) by Md(x, y, t) = t t + d(x, y) . Then (X, Md, ·) is a GV-fuzzy quasi-metric space called standard fuzzy quasi- metric space and (Md, ·) is the fuzzy quasi-metric induced by d. Furthermore, it is easy to check that (Md) −1 = Md−1 and (Md) i = Mds. Finally, from Proposition 2.8 and Remark 2.15, it follows that the topology τd, generated by d, coincides with the topology τMd generated by the induced fuzzy quasi-metric (Md, ·). Definition 2.17. We say that a topological space (X, τ) admits a compatible KM (resp. GV)-fuzzy quasi-metric if there is a KM (resp. GV)-fuzzy quasi- metric (M, ∗) on X such that τ = τM . It follows from Example 2.16 that every quasi-metrizable topological space admits a compatible GV-fuzzy quasi-metric. In Section 3 we shall establish that, conversely, the topology generated by a KM-fuzzy quasi-metric space is quasi-metrizable. 3. Quasi-metrizability of the topology of a fuzzy quasi-metric space A slight modification of the proof of Theorem 1 of [6], permits us to show the following result. Lemma 3.1. Let (X, M, ∗) be a KM-fuzzy quasi-metric space. Then {Un : n=2, 3, ...} is a base for a quasi-uniformity UM on X compatible with τM , where Un = {(x, y) ∈ X × X : M(x, y, 1/n) > 1 − 1/n}, for n = 2, 3, ... Moreover the conjugate quasi-uniformity (UM ) −1 coincides with UM−1 and it is compatible with τM−1. From Example 2.16, Lemma 3.1 and the well-known result that the topolo- gy generated by a quasi-uniformity with a countable base is quasi-pseudo- metrizable ([2]), we immediately deduce the following. Theorem 3.2. For a topological space (X, τ) the following are equivalent. (1) (X, τ) is quasi-metrizable. (2) (X, τ) admits a compatible GV-fuzzy quasi-metric. (3) (X, τ) admits a compatible KM-fuzzy quasi-metric. Remark 3.3. It is almost obvious that the uniformity UMi coincides with the uniformity (UM ) s := UM ∨ (UM ) −1 134 V. Gregori and S. Romaguera 4. Bicomplete fuzzy quasi-metric spaces There exist many different notions of quasi-uniform and quasi-metric com- pleteness in the literature (see [10]). Then, by Lemma 3.1 and Remark 3.3, one can define in a natural way the corresponding notions of completeness in a fuzzy setting and easily deduce several properties taking into account the well-known completeness properties of quasi-uniform and quasi-metric spaces (compare with [6], where these ideas are used to study completeness in the fuzzy metric case). In this section we only consider the notion of bicompleteness because it provides a satisfactory theory of quasi-uniform and quasi-metric completeness. Let us recall that a quasi-metric space (X, d) is bicomplete provided that (X, ds) is a complete metric space. In this case we say that d is a bicomplete quasi-metric on X. A metrizable topological space (X, τ) is said to be completely metrizable if it admits a compatible complete metric. On the other hand, a fuzzy metric space (X, M, ∗) is called complete ([5]) if every Cauchy sequence is convergent, where a sequence (xn)n is Cauchy provided that for each r ∈ (0, 1) and each t > 0, there exists an n0 such that M(xn, xm, t) > 1 − r for every n, m ≥ n0. If (X, M, ∗) is a complete fuzzy metric space, we say that (M, ∗) is a complete fuzzy metric on X. It was proved in [6] that a topological space is completely metrizable if and only if it admits a compatible complete fuzzy metric. Definition 4.1. A KM (resp. GV)-fuzzy quasi-metric space (X, M, ∗) is called bicomplete if (X, Mi, ∗) is a complete fuzzy metric space. In this case, we say that (M, ∗) is a bicomplete KM (resp. GV)-fuzzy quasi-metric on X. Proposition 4.2. (a) Let (X, M, ∗) be a bicomplete KM-fuzzy quasi-metric space. Then (X, τM ) admits a compatible bicomplete quasi-metric. (b) Let (X, d) be a bicomplete quasi-metric space. Then (X, Md, ·) is a bicomplete GV-fuzzy quasi-metric space. Proof. (a) Let d be a quasi-metric on X inducing the quasi-uniformity UM. Then d is compatible with τM . Now let (xn)n be a Cauchy sequence in (X, d s). Clearly (xn)n is a Cauchy sequence in the fuzzy metric space (X, M i, ∗). So it converges to a point y ∈ X with respect to τMi . Hence (xn)n converges to y with respect to τds. Consequently d is bicomplete. (b) This part is almost obvious because (Md) i = Mds (see Example 2.16), and thus each Cauchy sequence in (X, (Md) i, ·) is clearly a Cauchy sequence in (X, ds). � Extending the classical metric theorem, it was independently proved in [1] and [12], that every quasi-metric space admits a (quasi-metric) bicompletion which is unique up to isometry. Although the problem of completion of fuzzy metric spaces in the sense of Kramosil and Michalek has a satisfactory solution Fuzzy quasi-metric spaces 135 ([14]), the corresponding situation for fuzzy metric spaces in the sense of George and Veeramani is quite different. In fact, it was obtained in [7] an example of a fuzzy metric space (X, M, ∗) that does not admit completion, i.e. there no exist any complete fuzzy metric space having a dense subspace isometric to (X, M, ∗). A characterization of those fuzzy metric spaces (in the sense of George and Veeramani) that admit a fuzzy metric completion has recently been obtained in [8]. Although the problem of bicompletion for GV-fuzzy quasi-metric spaces will be discussed elsewhere, we next present some concepts and facts that are basic in solving this problem. Definition 4.3. Let (X, M, ∗) and (Y, N, ⋆) be two KM (resp. GV)-fuzzy quasi- metric spaces. Then (a) A mapping f from X to Y is called an isometry if for each x, y ∈ X and each t > 0, M(x, y, t) = N(f(x), f(y), t). (b) (X, M, ∗) and (Y, N, ⋆) are called isometric if there is an isometry from X onto Y. Definition 4.4. Let (X, M, ∗) be a KM (resp. GV)-fuzzy quasi-metric space. A KM (resp. GV)-fuzzy quasi-metric bicompletion of (X, M, ∗) is a bicom- plete KM (resp. GV)-fuzzy quasi-metric space (Y, N, ⋆) such that (X, M, ∗) is isometric to a τNi-dense subspace of Y . Proposition 4.5. áLet (X, M, ∗) be a KM-fuzzy quasi-metric space and (Y, N, ⋆) a bicomplete KM-fuzzy quasi-metric space. If there is a τMi-dense subset A of X and an isometry f : (A, M, ∗) → (Y, N, ⋆), then there exists a unique isom- etry F : (X, M, ∗) → (Y, N, ⋆) such that F |A= f. Proof. áIt is clear that f is a quasi-uniformly continuous mapping from the quasi-uniform space (A, UM |A×A) to the quasi-uniform space (Y, UN). By Theorem 3.29 of [2], f has a unique quasi-uniformly continuous extension F : (X, UM) → (Y, UN). We shall show that actually F is an isometry from (X, M, ∗) to (Y, N, ⋆). Indeed, let x, y ∈ X and t > 0. Then, there exist two sequences (xn)n and (yn)n in A such that xn → x and yn → y with respect to τMi. Thus F(xn) → F(x) and F(yn) → F(y) with respect to τNi. Choose ε ∈ (0, 1) with ε < t. Therefore, there is nε such that for n ≥ nε, M(x, xn, ε/2) > 1 − ε, M(yn, y, ε/2) > 1 − ε, N(F(xn), F(x), ε/2) > 1 − ε, N(F(y), F(yn), ε/2) > 1 − ε. Thus M(x, y, t) ≥ M(x, xn, ε/2) ∗ M(xn, yn, t − ε) ∗ M(yn, y, ε/2) ≥ (1 − ε) ∗ N(F(xn), F(yn), t − ε) ∗ (1 − ε) ≥ (1 − ε) ∗ [(1 − ε) ⋆ N(F(x), F(y), t − 2ε) ⋆ (1 − ε)] ∗ (1 − ε). By continuity of ∗ and ⋆ and by left continuity of N(F(x), F(y), ) it follows that M(x, y, t) ≥ N(F(x), F(y), t). Similarly we show that N(F(x), F(y), t) ≥ M(x, y, t). Consequently F is an isometry from (X, M, ∗) to (Y, N, ⋆). � 136 V. Gregori and S. Romaguera Corollary 4.6. Let (X, M, ∗) be a GV-fuzzy quasi-metric space and (Y, N, ⋆) a bicomplete GV-fuzzy quasi-metric space. If there is a τMi-dense subset A of X and an isometry f : (A, M, ∗) → (Y, N, ⋆), then there exists a unique isometry F : (X, M, ∗) → (Y, N, ⋆) such that F |A= f. References [1] A. Di Concilio, Spazi quasimetrici e topologie ad essi associate, Rend. Accad. Sci. Fis. Mat. Napoli 38 (1971), 113-130. [2] P. Fletcher, W.F. Lindgren, Quasi-Uniform Spaces, Marcel Dekker, New York, 1982. [3] A. George, P. Veeramani, On some results in fuzzy metric spaces, Fuzzy Sets and Sys- tems 64 (1994), 395-399. [4] A. George, P. Veeramani, Some theorems in fuzzy metric spaces, J. Fuzzy Math. 3 (1995), 933-940. [5] A. George, P. Veeramani, On some results of analysis for fuzzy metric spaces, Fuzzy Sets and Systems 90 (1997), 365-368. [6] V. Gregori, S. Romaguera, Some properties of fuzzy metric spaces, Fuzzy Sets and Systems 115 (2000), 485-489. [7] V. Gregori, S. Romaguera, On completion of fuzzy metric spaces, Fuzzy Sets and Sys- tems 130 (2002), 399-404. [8] V. Gregori, S. Romaguera, Characterzing completable fuzzy metric spaces, Fuzzy Sets and Systems, to appear. [9] I. Kramosil, J. Michalek, Fuzzy metric and statistical metric spaces, Kybernetika 11 (1975), 326-334. [10] H.P.A. Künzi, Nonsymmetric distances and their associated topologies: About the ori- gins of basic ideas in the area of asymmetric topology, in: Handbook of the History of General Topology (eds. C.E. Aull and R. Lowen), vol. 3, Kluwer (Dordrecht, 2001), pp. 853-968. [11] D. Mihet, A Banach contraction theorem in fuzzy metric spaces, Fuzzy Sets and Sys- tems, to appear. [12] S. Salbany, Bitopological Spaces, Compactifications and Completions, Math. Mono- graphs, no. 1, Dept. Math. Univ. Cape Town, 1974. [13] B. Schweizer, A. Sklar, Statistical metric spaces, Pacific J. Math. 10 (1960), 314-334. [14] H. Sherwood, On the completion of probabilistic metric spaces, Z. Wahrsch. verw. Geb. 6 (1966), 62-64. Received February 2003 Accepted June 2003 V. Gregori (vgregori@mat.upv.es) Escuela Politécnica Superior de Gandia, Universidad Politécnica de Valencia, 46730 Grau de Gandia, Valencia, Spain. S. Romaguera (sromague@mat.upv.es) Escuela de Caminos, Departamento de Matemática Aplicada, Universidad Politécnica de Valencia, 46071 Valencia, Spain.