On I-topology generated by fuzzy norm

On I-topology generated by fuzzy norm

Fuzzy Sets and Systems 157 (2006) 2739 – 2750 www.elsevier.com/locate/fss On I-topology generated by fuzzy norm夡 Jin-xuan Fang Department of Mathemat...

248KB Sizes 0 Downloads 4 Views

Fuzzy Sets and Systems 157 (2006) 2739 – 2750 www.elsevier.com/locate/fss

On I-topology generated by fuzzy norm夡 Jin-xuan Fang Department of Mathematics, Nanjing Normal University, Nanjing, Jiangsu 210097, China Received 5 February 2005; received in revised form 25 March 2006; accepted 27 March 2006 Available online 27 April 2006

Abstract In this paper, we point out that an I-topology T· on the fuzzy normed linear space (X,  · , min, max) constructed by Das and Das [Fuzzy topology generated by fuzzy norm, Fuzzy Sets and Systems 107 (1999) 349–354] is incompatible with the linear structure on X, that is, (X,  · , min, max) is not an I-topological vector space with respect to the I-topology T· . Therefore, we construct a new I-topology T∗· on the fuzzy normed linear space (X,  · , L, R) by using fuzzy norm  · . We study some of its properties and prove that if R  max, then (X,  · , L, R) is a Hausdorff locally convex I-topological vector space with respect to the I-topology T∗· . In addition, we also study the relations among three I-topologies T∗· , T· and (), where () is the induced I-topology of the crisp vector topology  determined by fuzzy norm  · . © 2006 Elsevier B.V. All rights reserved. Keywords: Fuzzy normed linear space; I-topology; Induced I-topology; I-topological vector space; Locally convex I-topological vector space

1. Introduction Kaleva and Seikkala [9] introduced and studied fuzzy metric spaces. Inspired by this work, Felbin [6] introduced the concept of fuzzy normed linear space (FNS). Xiao and Zhu [14] studied the linear topological structure of FNSs. Das and Das [2] constructed a fuzzy topology T· on the FNS (X,  · , min, max) and studied some basic properties of this fuzzy topology. According to the standardized terminology in [8], the fuzzy topology related in [2] should be called I-topology, where I = [0, 1]. It is well known that an ordinary normed linear space is necessarily a topological vector space for the topology determined by the norm. This leads naturally to the following question: is the FNS (X,  · , min, max) an I-topological vector space with respect to the I-topology T· constructed by Das and Das? In this paper, we first give a negative answer to this question, and so a shortcoming of the I-topology T· constructed ∗ on the FNS (X,  · , L, R) and study some of by Das and Das is disclosed. Next, we introduce a new I-topology T· its properties. We prove that under the condition of R  max, the FNS (X,  · , L, R) is a Hausdorff locally convex ∗ . Finally, the relations among three I-topologies T ∗ , T I-topological vector space with respect to the I-topology T· · · and () on the FNS (X,  · , L, R) are discussed, where () is the induced I-topology of the crisp vector topology ∗ ⊂ () ⊂ T , and give  determined by fuzzy norm  ·  (see [5,14]). We prove the following proper inclusions: T· · ∗ a necessary and sufficient condition for the equality T· = () to hold. 夡

Project is supported by the Natural Science Foundation of Educational Department of Jiangsu Province of China (04KJB110061). E-mail address: [email protected]

0165-0114/$ - see front matter © 2006 Elsevier B.V. All rights reserved. doi:10.1016/j.fss.2006.03.024

2740

J.-x. Fang / Fuzzy Sets and Systems 157 (2006) 2739 – 2750

2. Preliminaries Throughout this paper, let R = (−∞, +∞), R+ = [0, +∞) and I = [0, 1], and let N be the set of all positive integers. We denote the family of all fuzzy sets on X by I X . For r ∈ [0, 1], we denote a fuzzy set which takes the constant value r on X by r. A fuzzy point x on X (0 <  1) is a fuzzy set defined by   if y = x, x (y) = 0 otherwise. A crisp point x in X may regard as a fuzzy point x1 . Let A ∈ I X and x be a fuzzy point. We say x quasi-coincides with A, denoted by x ∈A, if A(x) > 1 − ; say x belongs to A, denoted by x ∈ A, if A(x) . We do not use the definition of a fuzzy real number in [7]. A mapping  : R → I is called a fuzzy set of real numbers. A fuzzy set of real numbers  is called non-negative if (t) = 0 for t < 0. The set of all non-negative upper semi-continuous normal convex fuzzy sets of real numbers is denoted by G+ . Obviously, every  in G+ is a fuzzy + − interval [3], i.e., for every  ∈ (0, 1], the -level set [] = {t ∈ R | (t) } is a closed interval [−  ,  ] or [ , +∞). Definition 1 (Felbin [6]). Let X be a vector space over R,  ·  : X → G+ and let the mappings L, R : I × I → I be symmetric, non-decreasing in both arguments and satisfy L(0, 0) = 0 and R(1, 1) = 1. Write [x] = [|x|]1 , |x|2 ]

for x ∈ X,  ∈ (0, 1]

and suppose for all x ∈ X, x  = , there exists 0 ∈ (0, 1] independent of x such that for all  0 , (A) |x|2 < +∞. The quadruple (X,  · , L, R) is called a fuzzy normed linear space (FNS) and  ·  a fuzzy norm, if the following conditions are satisfied: (FN-1) x = 0˜ if and only if x = ; (FN-2) rx = |r| x, x ∈ X, r ∈ R; (FN-3) for all x, y ∈ X, (a) whenever s |x|11 , t |y|11 and s + t |x + y|11 , x + y(s + t)L(x(s), y(t)), (b) whenever s |x|11 , t |y|11 and s + t |x + y|11 , x + y(s + t)R(x(s), y(t)). Remark 1. There is a little difference between the above definition and the original definition of FNS given in [6]. In [6], other than condition (A), the definition also requires that  ·  : X → G+ to satisfy the condition (B) inf |x|1 > 0 for all x  = . 

Lemma 1 (Fang and Song [5]). Let (X,  · , L, R) be a FNS, where R satisfies the condition lima→0 R(a, a) = 0. Then there exists a topology  on X such that (X, ) is a metrizable Hausdorff topological vector space having U = {Uε, : ε > 0,  ∈ (0, 1]} as a neighborhood base of , where Uε, = {x ∈ X : |x|2 < ε}.

(2.1)

 is called the vector topology determined by fuzzy norm  · . Lemma 2 (Xiao and Zhu [14])). Let (X,  · , L, R) be a FNS. Then (1) R  max implies that for every  ∈ (0, 1], |x + y|2 |x|2 + |y|2 ,

∀ x, y ∈ X;

J.-x. Fang / Fuzzy Sets and Systems 157 (2006) 2739 – 2750

2741

(2) lima→0 R(a, a) = 0 implies that for every  ∈ (0, 1], there exists  = () ∈ (0, ] such that 



|x + y|2 |x|2 + |y|2 ,

∀ x, y ∈ X.

Definition 2 (Das and Das [2]). Let (X,  · , L, R) be a FNS. For fixed x ∈ X,  ∈ (0, 1] and ε > 0, the fuzzy set  (x, ε) on X given by   (x, ε)(y) =

 if |y − x|2 < ε, 0 otherwise

(2.2)

is said to be an -sphere in (X,  · , L, R). Remark 2. By (2.1) and (2.2), it is easy to verify  (x, ε) = (x + Uε, ) ∩  = x + Uε, ∩ .

(2.3)

Definition 3 (Das and Das [2]). A fuzzy set  on (X,  · , L, R) is said to be  · -open if for every x ∈ supp  = {x ∈ X | (x) > 0}, there exist ε > 0 and  ∈ (0, 1] such that  (x, ε) ⊂ . Lemma 3 (Das and Das [2]). Let (X,  · , L, R) be a FNS. Then the collection T· = { ∈ I X |  is  · -open} is an I-topology [1] on X. For convenience, we call it the Das’s I-topology. Remark 3. It is easy to see that T· is a stratified I-topology [11] (namely, an I-topology in the sense of Lowen [12]). Lemma 4 (Das and Das [2]). In the FNS (X,  · , min, max), every -sphere  (x, ε) is an I-open fuzzy set for T· , i.e.,  (x, ε) ∈ T· . Lemma 5. Let 0 <  < 1 and  (x0 , ε) be an -sphere in the FNS (X,  · , min, max). Then for any given  ∈ (, 1], the fuzzy set  on X defined by  (x) =

 if x = x0 ,  (x0 , ε)(x) if x  = x0

is I-open for T· and  (x0 , ε) ⊂ . Proof. By Lemma 4,  (x0 , ε) ∈ T· . Besides, it is easy to see that  (x0 , ε) ⊂  and supp  = supp  (x0 , ε). Thus, it follows from Definition 3 that  ∈ T· .  Let X be a vector space over the field K (K = R or C), A, B ∈ I X and t ∈ K. Then A + B and tA are defined by (A + B)(x) = sup (A(u) ∧ B(v)) u+v=x

and (tA)(x) = A(x/t),  (0A)(x) =

t  = 0,

sup A(y) if x = ,

y∈X

0

if x  = ,

2742

J.-x. Fang / Fuzzy Sets and Systems 157 (2006) 2739 – 2750

respectively. In particular, for the fuzzy points x , y , we have (x + A)(y) =  ∧ A(y − x), x + y = (x + y)∧ ,

(x + A)(y) = A(y − x);

tx = (tx) .

Definition 4 (Katsaras [10]). A stratified I-topology T on a vector space X is said to be an I-vector topology, if the following two mappings f : X × X → X, (x, y) → x + y, g : K × X → X,

(t, x) → tx,

are continuous, where K is equipped with the I-topology induced by the usual topology, and X × X and K × X are equipped with the corresponding product I-topologies. A vector space X with an I-vector topology T , denoted by (X, T ), is called an I-topological vector space (I-TVS). Definition 5 (Liu and Lao [11]). Let (X, ) be an I-topological space and x ∈ Pt(I X ). (1) A fuzzy set U on X is called Q-neighborhood of x iff there exists G ∈  such that x ∈ G ⊂ U. (2) A family Ux of Q-neighborhoods of x is called a Q-neighborhood base of x iff for every Q-neighborhood A of x , there exists U ∈ Ux such that U ⊂ A. Definition 6 (Wu and Fang [13]). An I-topological vector space (X, T ) is said to be of QL-type, if there exists a family U of fuzzy sets on X such that for each  ∈ (0, 1], U = {U ∩ r | U ∈ U, r ∈ (1 − , 1]} is a Q-neighborhood base of  in (X, T ). The family U is called a Q-prebase for T . Lemma 6 (Fang [4]). Let T be a stratified I-topology on a vector space X. Then (1) the mapping f (addition) is continuous iff for every fuzzy point (x, y) in X × X and every Q-neighborhood W of (x + y) , there exist a Q-neighborhood U of x and a Q-neighborhood V of y such that U + V ⊂ W ; (2) the mapping g (scalar multiplication) is continuous iff for every fuzzy point (t, x) in K × X and every Q-neighborhood W of (tx) , there exist a Q-neighborhood U of x and > 0 such that sU ⊂ W for all s ∈ K with |s − t| < . Lemma 7 (Fang [4]). Let (X, T ) be an I-TVS and U a Q-neighborhood base of  in X,  ∈ (0, 1]. Then the following conclusions hold: (1) If U ∈ U or U = r, where r ∈ (1 − , 1], then there exists 0 ∈ (0, ) such that for each  ∈ [0 , 1] there is a V ∈ U such that V ⊂ U . (2) If U, V ∈ U , then there exists W ∈ U such that W ⊂ U ∩ V . (3) If U ∈ U , then there exists V ∈ U such that V + V ⊂ U . (4) If U ∈ U , then there exists V ∈ U such that tV ⊂ U for all t ∈ K with |t| 1. ˜ . (5) If U ∈ U and x ∈ X, there exists  > 0 such that x ∈U Conversely, let X be a vector space over K such that every  ∈ (0, 1] has a family U of fuzzy sets on X satisfying the conditions (1)–(5), then there exists a unique I-topology T on X such that (X, T ) is an I-TVS and U is a Q-neighborhood base of  . Definition 7. An I-TVS (X, T ) is said to be locally convex, if for each  ∈ (0, 1], there is a base of Q-neighborhoods of  consisting of convex fuzzy sets.

J.-x. Fang / Fuzzy Sets and Systems 157 (2006) 2739 – 2750

2743

3. A shortcoming of Das’s I-topology Theorem 1. In the FNS (X,  · , min, max), (1) the addition, i.e., the mapping f : (X, T· ) × (X, T· ), (x, y) → x + y, is continuous; however, (2) the scalar multiplication, i.e., the mapping g : R × (X, T· ), (t, x) → tx, is not continuous. Proof. (1) Suppose that W is a Q-neighborhood of (x + y) , then there exists  ∈ T· such that (x + y) ∈  ⊂ W . By Definition 3, there exist ε > 0 and  ∈ (0, 1] such that  (x + y, ε) ⊂  ⊂ W . By conclusion (1) of Lemma 2 , it is not difficult to verify that Uε/2, + Uε/2, ⊂ Uε, , and so we have (x + Uε/2, ) + (y + Uε/2, ) ⊂ (x + y) + Uε, , from which it follows that (x + Uε/2, ) ∩  + (y + Uε/2, ) ∩  ⊂ [(x + y) + Uε, ] ∩ . By Remark 2, we obtain ε ε  (x, ) +  (y, ) ⊂  (x + y, ε) ⊂ W. 2 2

(3.1)

If  > 1 − , then it is easy to see that  (x, ε/2),  (y, ε/2) are the T· -open Q-neighborhoods of x and y , respectively. Thus, by (3.1) and Lemma 6 we know that the addition is continuous. If 1 − , we define the fuzzy sets A and B on X as follows:   if z = x, A(z) =  (x, 2ε )(z) if z  = x and

 B(z) =

 if z = y,  (y, 2ε )(z) if z  = y,

where  = (x + y) > 1 − . We can prove that A + B ⊂ W.

(3.2)

In fact, when z = x + y, we have (A + B)(z) =  = (z); when z  = x + y, we have (A + B)(z) = [A(x) ∧ B(z − x)] ∨ [A(z − y) ∧ B(y)] ∨ ε ε =  (y, )(z − x) ∨  (x, )(z − y) ∨ 2 2

sup u+v=z

[A(u) ∧ B(v)]

u=x, v=y

sup u+v=z

u=x, v=y

ε ε [ (x, )(u) ∧  (y, )(v)], 2 2

and so if |z − (x + y)|2 < ε, then (A + B)(z)  =  (x + y, ε)(z)(z); if |z − (x + y)|2 ε, then we have |u−x|2 ε/2 or |v −y|2 ε/2, where z = u+v. Hence (A+B)(z) = 0 (z). This shows that A+B ⊂  ⊂ W , so (3.2) holds. By Lemma 5, we know that A and B are the T· -open Q-neighborhoods of x and y , respectively. By (3.2) and conclusion (1) of Lemma 6, we know that the addition is continuous. (2) Let (t0 , x) ∈ R × X, x  =  and  ∈ (0, 1]. Taking ,  ∈ (0, 1] such that  < 1 −  < , we define a fuzzy set  on X by   if z = t0 x, (z) =  if z  = t0 x. Obviously, for every y ∈ supp  and ε > 0,  (y, ε) ⊂  and (t0 x) =  > 1 − . Hence  is a T· -open Q-neighborhood of t0 x . However, it is not difficult to prove that for every Q-neighborhood U of x and every > 0, (t0 + )U ⊂ / .

(3.3)

2744

J.-x. Fang / Fuzzy Sets and Systems 157 (2006) 2739 – 2750

In fact, when z = (t0 + )x, we have [(t0 + )U ](z) = U (x) > 1 − ; but, (z) =  < 1 − . So, (3.3) holds. From (2) of Lemma 6, we know that the scalar multiplication is not continuous.  Remark 4. From conclusion (2) of Theorem 1, we know that the FNS (X,  · , min, max) is not an I-topological vector space with respect to Das’s I-topology T· . This is a shortcoming of Das’s I-topology. 4. I -linear topological structure of a fuzzy linear normed space By modifying the definition of  · -open fuzzy set, we give the following: Definition 8. A fuzzy set  on (X,  · , L, R) is said to be  · -linearly open if for every x ∈ supp  and  ∈ (0, (x)), there exists ε > 0 such that  (x, ε) ⊂ . Theorem 2. Let (X,  · , L, R) be a FNS. Then the collection ∗ = { ∈ I X |  is  · -linearly open} T·

is a stratified I-topology on (X,  · , L, R), and it is called the I-topology generated by the fuzzy norm. ∗ for Proof. (i) Let r ∈ (0, 1]. For every x ∈ supp r = X,  ∈ (0, r) and any ε > 0, we have  (x, ε) ⊂ r. So r ∈ T· all r ∈ (0, 1]. Again supp 0 = ∅, So 0 is also  · -linearly open. ∗ and ( ∩  )(x) =  > 0. Then for every  ∈ (0, ), we have  ∈ (0,  (x)) (i = 1, 2). Since (ii) Let 1 , 2 ∈ T· 1 2 i ∗ i ∈ T· , there exists εi > 0 such that  (x, εi ) ⊂ i (i = 1, 2). Let ε = min{ε1 , ε2 }, we have

 (x, ε) ⊂  (x, ε1 ) ∩  (x, ε2 ) ⊂ 1 ∩ 2 . ∗ . Hence 1 ∩ 2 ∈ T·  ∗ and ( (iii) Let {i }i∈ ⊂ T· i ∈ such that i∈ i )(x) =  > 0. Then for every  ∈ (0, ),there exists some  0 ∗ ∗ .  ∈ (0, i0 (x)). Since i0 ∈ T· , there exists ε0 > 0 such that  (x, ε0 ) ⊂ i0 ⊂ i∈ i . Hence i∈ i ∈ T· ∗ Therefore, T· is a stratified I-topology on (X,  · , L, R). 

Theorem 3. Let R satisfy the condition R  max. Then every -sphere 0 (x0 , ε) in the FNS (X,  · , L, R) is  · ∗ . linearly open, i.e. 0 (x0 , ε) ∈ T· Proof. Let x ∈ supp 0 (x0 , ε), then |x0 − x|20 < ε and 0 (x0 , ε)(x) = 0 . For every  ∈ (0, 0 ), let = ε − |x0 − x|20 . It is not difficult to prove that  (x, ) ⊂ 0 (x0 , ε). In fact, if |x − y|2 < , by Lemma 2 we have |x0 − y|20  |x0 − x|20 + |x − y|20  |x0 − x|20 + |x − y|2 < |x0 − x|20 + = ε.

Hence,  (x, )(y) =  < 0 = 0 (x0 , ε)(y). Again, if |x − y|2  , then  (x, )(y) = 0 0 (x0 , ε)(y). Therefore,  (x, ) ⊂ 0 (x0 , ε). This shows that 0 (x0 , ε) is  · -linearly open.  ∗ ) is a locally convex Theorem 4. Let (X,  · , L, R) be a FNS, where R satisfies the condition R  max. Then (X, T· I-TVS and for every  ∈ (0, 1],

U = {Uε, ∩  | ε > 0,  ∈ (1 − , 1]} = { (, ε) | ε > 0,  ∈ (1 − , 1]} is a Q-neighborhood base of  , where Uε, is defined by (2.1). ∗ ) is a Hausdorff I-TVS. Also, if (X,  · , L, R) satisfies condition (B), i.e., inf  |x|1 > 0 for all x  = , then (X, T· Proof. First, we verify that U ( ∈ (0, 1]) satisfies conditions (1)–(5) of Lemma 7. (1) Let U = Uε, ∩  ∈ U or U =  with  ∈ (1 − , 1], then there exists 0 ∈ (0, ) such that  > 1 − 0 . So for each  ∈ [0 , 1] we have  > 1 − . This shows that Uε, ∩  ∈ U . Taking V = Uε, ∩ , we get V ⊂ U .

J.-x. Fang / Fuzzy Sets and Systems 157 (2006) 2739 – 2750

2745

(2) Let Uεi ,i ∩ i ∈ U (i = 1, 2). Taking ε = min{ε1 , ε2 } and  = min{1 , 2 }, then we have Uε, ∈ U and Uε, ∩  ⊂ (Uε1 ,1 ∩ 1 ) ∩ (Uε2 ,2 ∩ 2 ). (3) Let Uε, ∩  ∈ U . Since R  max, by (1) of Lemma 2 we have |x + y|2 |x|2 + |y|2 ,

∀ x, y ∈ X.

(4.1)

It follows from (4.1) that Uε/2, + Uε/2, ⊂ Uε, . Obviously, Uε/2, ∩  ∈ U and Uε/2, ∩  + Uε/2, ∩  ⊂ Uε, ∩ . (4) If Uε, ∩  ∈ U , then it is easy to see that Uε, is a balanced set, i.e., tUε, ⊂ Uε, for all t ∈ R with |t| 1 and then we have t (Uε, ∩ ) = tUε, ∩  ⊂ Uε, ∩ . (5) If Uε, ∩  ∈ U , then  > 1 − . For every x ∈ X, taking t > |x|2 /ε, we have |(1/t)x|2 = (1/t)|x|2 < ε, ∼

which implies that (1/t)x ∈ Uε, , and so (Uε, ∩ )((1/t)x) = 1 ∧  =  > 1 − . i.e., x ∈ t (Uε, ∩ ). Thus, from the second part of Lemma 7, there exists a unique I-topology T on X such that (X, T ) is an I-TVS and U is a Q-neighborhood base of  . In addition, it is easy to verify that for each r ∈ (0, ], [ (, ε)]r = Uε, and Uε, is convex. Hence every  (, ε) in U is a convex fuzzy set. Therefore, (X, T ) is locally convex. ∼ ∗ . If  ∈ T ∗ and x ∈ , i.e., (x) > 1 − , then for every  ∈ (1 − , (x)), there exists Now we prove T = T·  · ε > 0 such that  (x, ε) ⊂ . Notice that  (x, ε) = x + Uε, ∩  is a Q-neighborhood of x for T , hence  is also ∗ ⊂ T . On the other hand, if  ∈ T and x ∈ supp , for every that of x for T . This shows that  ∈ T , and so T· ∼

 ∈ (0, (x)), let  = 1 − , then we have x ∈ . Thus there exists Uε, ∩  ∈ U such that x + Uε, ∩  ⊂ , i.e., ∗ , and so T ⊂ T ∗ . Therefore T ∗ = T .  (x, ε) ⊂ . This shows that  ∈ T· · · ∗ ) is Hausdorff. Finally, we prove that if (X,  · , L, R) satisfies condition (B), then (X, T· Let x and y be two fuzzy points in X and x  = y. By condition (B) of Definition 1, we know that inf  |x − y|1 = ∗ and  (y, /2) is a Q-neighborhood of y for T ∗ . > 0. Obviously, 1 (x, /2) is a Q-neighborhood of x for T· 1 · Now we prove that 1 (x, /2) ∩ 1 (y, /2) = ∅. If not, then there exists z ∈ X such that 1 (x, /2)(z) > 0 and 1 (y, /2)(z) > 0, i.e. |x − z|12 < /2 and |z − y|12 < /2. Thus, we have |x − y|11 |x − y|12 |x − z|12 + |z − y|12 < /2 + /2 = , ∗ ) is Hausdorff. which is a contradiction. Therefore, (X, T·



Using Theorem 4, we can obtain the following: (n)

Corollary 1. Let (X,  · , L, R) be a FNS, where R satisfies the condition R  max, and let {x } be a sequence n (n)

∗ T·

of fuzzy points in X and x be a fuzzy point in X. Then x −→ x if and only if limn→∞ |x (n) − x|2 = 0 for all n  ∈ (1 − , 1] and limn→∞ n . Corollary 2. Let (X,  · , L, R) be a FNS, where R satisfies the condition R  max, and let {xn } be a sequence of ∗ T·

crisp points in X and x be a crisp point in X. Then xn −→ x if and only if limn→∞ |xn − x|2 = 0 for all  ∈ (0, 1], 

i.e., xn −→ x. 5. Relations among three I-topologies on fuzzy normed linear space ∗ and () on a FNS (X,  · , L, R), where In this section, we discuss the relations among three I-topologies T· , T· ∗ T· is Das’s I-topology, T· is the I-topology generated by fuzzy norm and () is the induced I-topology by the crisp vector topology  determined by fuzzy norm. ∗ ⊂ () ⊂ T . Theorem 5. Let (X,  · , L, R) be a FNS, where R satisfies lima→0 R(a, a) = 0. Then T· ·

2746

J.-x. Fang / Fuzzy Sets and Systems 157 (2006) 2739 – 2750

∗ ⊂ (), we need only to prove that  ∈ T ∗ implies that  ∈ (), i.e., () = {x ∈ X | Proof. To prove T·  · ∗ , there exists ε > 0 (x) > } ∈  for all  ∈ [0, 1). Let x ∈  (), i.e., (x) > . Take  ∈ (, (x)). By  ∈ T· such that  (x, ε) ⊂ . Since  (x, ε) = (x + Uε, ) ∩ , we have x + Uε, ⊂  ( (x, ε)) ⊂  (). This shows that  () is a neighborhood of x in (X, ), and so  () ∈ . Hence  ∈ (). Let  ∈ () and (x) > 0. Take any r ∈ (0, (x)). We have x ∈ r () ∈ . So, by Lemma 1, there exist ε > 0 and  ∈ (0, 1] such that x + Uε, ⊂ r (). Set  = min{r, }, it follows that

 (x, ε) = (x + Uε, ) ∩  ⊂ (x + Uε, ) ∩ r ⊂ . Hence  ∈ T· , which shows that () ⊂ T· .



∗ = () if and only if Theorem 6. Let (X,  · , L, R) be a FNS, where R satisfies the condition R  max, then T·



for any ,  ∈ (0, 1), | · |2 is equivalent to | · |2 , i.e., there exist C1 , C2 > 0 such that 

C1 |x|2 |x|2 C2 |x|2

∀ x ∈ X.

Proof. Note that the condition R  max implies lima→0 R(a, a) = 0. By Lemma 1, (X, ) is a crisp topological vector space, where  is the vector topology determined by the fuzzy norm  · . Thus, we know that (X, ()) is a (QL)-type I-TVS and for each  ∈ (0, 1], B = {Uε, ∩ r | ε > 0,  ∈ (0, 1], r ∈ (1 − , 1]} is a Q-neighborhood base of  in (X, ()). By Theorem 3, we know that for each  ∈ (0, 1], U = {Uε, ∩  | ε > 0,  ∈ (1 − , 1]} ∗ ). is a Q-neighborhood base of  in (X, T· ∗ Necessity: Let T· = () and ,  ∈ (0, 1),  > . We write  = 1 − . Then for r ∈ (, 1], U1, ∩ r ∈ B . Since ∗ = (), it is easy to know that U ∗ T· 1, ∩ r ∈ B is also a Q-neighborhood of  in (X, T· ). So, there exist ε > 0 and s ∈ (1 − , 1] such that Uε,s ∩ s ⊂ U1, ∩ r, from which it follows that

Uε,s =  (Uε,s ∩ s) ⊂  (U1, ∩ r) = U1, .

(5.1) 

Notice that (εx/|x|2 + 1/n) ∈ Uε, ⊂ Uε,s for all n ∈ N. It follows from (5.1) that |x|2 < (1/ε)(|x|2 +1/n), 

and so |x|2 (1/ε)|x|2 . Besides, since  > , we have 

|x|2 |x|2 (1/ε)|x|2

for all x ∈ X. 

Therefore | · |2 is equivalent to | · |2 . ∗ . Let  ∈ () and x ∈ supp . Then for every Sufficiency: By Theorem 5, we need only to prove () ⊂ T·  ∈ (0, (x)), we have x ∈  () ∈ . By Lemma 1, there exist ε > 0 and  ∈ (0, 1] such that x + Uε, ⊂  (). In the following, we split two cases to discuss: (1) If , then we have x + Uε, ⊂ x + Uε, ⊂  (), and so  (x, ε) = (x + Uε, ) ∩  ⊂ .

(5.2) 

(2) If  > , and notice that | · |2 is equivalent to | · |2 , then there exists m > 0 such that 



m|x|2 |x|2 |x|2 for all x ∈ X.

(5.3)

It follows from (5.3) that Umε, ⊂ Uε, ⊂ Uε, . So, we have x + Umε, ⊂ x + Uε, ⊂  (), which implies that  (x, mε) = (x + Umε, ) ∩  ⊂ . ∗ . This shows that () ⊂ T ∗ . From (5.2) and (5.4), we infer  ∈ T· ·

(5.4) 

J.-x. Fang / Fuzzy Sets and Systems 157 (2006) 2739 – 2750

2747

∗ is a proper subset of () and () is a proper In the following, we shall give two examples, which show that T· subset of T· . Let C(R) denote the set of all continuous real valued functions on R. For each  ∈ (0, 1], we define a function | · | : C(R) → [0, ∞) by

|x| =

max

t∈[−1/,1/]

|x(t)|.

Obviously, for every x ∈ C(R), 0 <  <  1 implies |x| |x| . Lemma 8. Let x ∈ C(R) with x  = , t > 0 and  ∈ (0, 1), we define x (t) by x (t) = sup{ ∈ (0, 1] | t |x| }. Then the following conclusions hold: (1) (2) (3) (4)

If t > |x| , then x (t). kx (t) = x (t/|k|) for all k  = 0. x+y (s + t) max{ x (s), y (t)}. x (t) is upper semi-continuous for t.

Proof. (1) and (2) are obvious. (3) By definition of x+y (s+t), for any ε > 0, there exists 0 > x+y (s+t)−ε such that s+t |x+y| 0 |x| 0 +|y| 0 , which implies that s |x| 0 or t |y| 0 . So, we have x (s)  0 or y (t)  0 . It follows that max{ x (s), y (t)}  0 > x+y (s + t) − ε. By the arbitrariness of ε, (3) is proved. (4) We need only to prove that for each  ∈ (0, 1], ( x ) = {t | x (t) } is a closed set. Let tn ∈ ( x ) with tn → t. Since x (tn ), there exists n >  − /2n such that tn |x| n |x|−/2n . Setting n =  − /2n, we have max

s∈[−1/ n ,1/ n ]

|x(s)|tn

(n = 1, 2, . . .).

So, there exists sn ∈ [−1/ n , 1/ n ] such that |x(sn )|tn (n = 1, 2, . . .). Since {sn } ⊂ [−2/, 2/], there exists a subsequence {sni } of {sn } such that sni → s0 . Notice that sni ∈ [−1/ ni , 1/ ni ] and ni → . Hence s0 ∈ [−1/, 1/]. It follows from |x(sni )| tni that |x(s0 )|t, i.e., |x| t, which implies x (t) , i.e., t ∈ ( x ) . Therefore ( x ) is a closed set.  Example 1. Let ,  ∈ (0, 1) with  > , n ∈ (0, ) with n  0. We define a mapping  ·  : C(R) → G+ as follows: (i) If |x|  = 0, then ⎧ 0 ⎪ ⎪ ⎨ 1 x(t) =  ⎪ ⎪ ⎩ x (t) (ii) If there exists some n0 ⎧ ⎨1 x(t) = n ⎩ 0

if if if if

t < 0, 0 t |x| , |x| < t |x| , t > |x| .

∈ N such that |x|n

0 −1

= 0 and |x|n  = 0 (0 = ), then 0

if t = 0, if t ∈ (|x|n−1 , |x|n ] (n = n0 , n0 + 1, . . .), otherwise.

(iii) If |x| = 0 for all ∈ (0, 1], i.e., x = , then  1 if t = 0, ˜ x(t) = i.e.,  = 0. 0 otherwise, ∗ is a proper subset of (). Then (C(R),  · , min, max) is a FNS, and T·

2748

J.-x. Fang / Fuzzy Sets and Systems 157 (2006) 2739 – 2750

In fact, (FN-1) holds obviously. ˜ Hence, 0 · x =  = 0˜ = |0| · x. (FN-2) It is easy to verify 0 · x = 0. Let k  = 0. In the case of (i), by Lemma 7, we have ⎧ 0 if t/|k| < 0 ⎪ ⎪ ⎨ 1 if t/|k| ∈ [0, |x| ] (|k| · x)(t) = x(t/|k|) =  if t/|k| ∈ (|x| , |x| ] ⎪ ⎪ ⎩ x (t/|k|) if t/|k| > |x| ⎧ 0 if t < 0 ⎪ ⎪ ⎨ 1 if t ∈ [0, |kx| ] = = k · x(t),  if t ∈ (|kx| , |kx| ] ⎪ ⎪ ⎩ kx (t) if t > |kx| and so k · x = |k| · x. In the case of (ii), since |x|n  = 0 ⇐⇒ |kx|n  = 0 (k  = 0), we have ⎧ ⎨ 1 if t/|k| = 0 (|k| · x)(t) = x(t/|k|) = n if t/|k| ∈ (|x|n−1 , |x|n ] (n = n0 , n0 + 1, . . .) ⎩ 0 otherwise ⎧ ⎨ 1 if t = 0 = n if t ∈ (|kx|n−1 , |kx|n ] (n = n0 , n0 + 1, . . .) ⎩ 0 otherwise = kx(t) and so k · x = |k| · x. In the case of (iii), clearly, k ·  =  = 0˜ = |k| · . (FN-3) Obviously, |x|11 = 0 for every x ∈ C(R), and so (a) of (FN-3) holds. (b) Let s |x|11 , t |y|11 and s + t |x + y|11 , i.e., s, t 0. We shall prove that x + y(s + t)  max{x(s), y(t)}. Clearly, we may assume that s, t > 0. (i) Let |x + y|  = 0. We have ⎧ if s + t ∈ (0, |x| ], ⎨1 if s + t ∈ (|x + y| , |x + y| ], x + y(s + t) =  ⎩ x+y (s + t) if s + t > |x + y| . Notice that 0 < s + t |x + y| ⇒ s |x| or t |y| ⇒ x(s) = 1 or y(t) = 1. Hence, when 0 < s + t |x + y| , we have x + y(s + t) = 1 = max{x(s), y(t)}. Notice that |x + y| < s + t |x + y| ⇒ s |x| or t |y| ⇒ x(s)  or y(t) . Hence, when |x + y| < s + t |x + y| , we have x + y(s + t) =  {x(s), y(t)}. When |x + y| < s + t, we consider the following two cases: Case 1: |x| < s and |y| < t. This implies x(s) x (s) and y(t)  y (t). Thus, by Lemma 7, we have x + y(s + t) = x+y (s + t) max{ x (s), y (t)} max{x(s), y(t)}. Case 2: |x| s or |y| t. This implies x(s) and y(t) . Thus, we have x + y(s + t) = x+y (s + t)  max{x(s), y(t)}. (ii) Let there exist some n0 ∈ N, such that |x + y|n

0 −1

= 0 and |x + y|n  = 0. 0

J.-x. Fang / Fuzzy Sets and Systems 157 (2006) 2739 – 2750

2749

Notice that s + t |x + y|n ⇒ s |x|n or t |y|n ⇒ x(s) n or y(t) n (n = n0 , n0 + 1, . . .). Hence,

when s + t ∈ |x + y|n−1 , |x + y|n , we have x + y(s + t) = n  max{x(s), y(t)},

n = n0 , n0 + 1, . . . .

(iii) Let x + y = . Obviously, x + y(s + t) max{x(s), y(t)}. (FN-3) is proved. Therefore (C(R), min, max) is a FNS.   Since |x|2 = |x| and |x|2 = |x| for all x ∈ C(R) with |x|  = 0, | · |2 is not equivalent to | · |2 . By ∗ Theorems 5 and 6, we can infer that T· is a proper subset of (). Example 2. Let us consider the FNS (R,  · , L, max), where  ·  : R → G+ is defined by  1 if t = |x|, x(t) = 0 otherwise.

(5.5)

We define a fuzzy set  on R by  1 if x ∈ [−1, 1], (x) = 1 2 otherwise. Then we can prove that  ∈ T· but  ∈ / (), and so () is a proper subset of T· by Theorem 5. In fact, for every  ∈ (0, 1], it follows from (5.5) that |x|2 = |x|, and so we have Uε, = (−ε, ε) for each ε > 0. This shows that the topology  determined by the fuzzy norm  ·  on R is the ordinary topology on R. Taking 0 <  21 and ε > 0, we have  (x, ε) ⊂ . Hence  ∈ T· . On the other hand, since 1/2 () = [−1, 1] ∈ / , we have ∈ / (). 6. Conclusions It is well known that every normed linear space is a topological vector space for the norm topology. However, the fuzzy normed linear space (X,  · , min, max) is not an I-topological vector space with respect to Das’s I-topology T· generated by fuzzy norm (see Theorem 1 and Remark 4). In order to get over this shortcoming, we have constructed a ∗ on the FNS (X,  · , L, R) and have shown that under the condition of R  max, (X,  · , L, R) new I-topology T· ∗ (see Theorems 2 and 4). is a Hausdorff locally convex I-topological vector space with respect to the I-topology T· ∗ We also have proved the following proper inclusions: T· ⊂ () ⊂ T· , where () is the induced I-topology of the crisp vector topology  determined by the fuzzy norm  ·  (see Theorem 5, Examples 1 and 2). Up to now, most of the researches about fuzzy normed linear spaces are based upon the crisp vector topology  determined by the fuzzy norm (see [5,6,14–16]). The present paper provides a new framework for studying theory of ∗ generated by the fuzzy norm. How to establish the corresponding theory FNSs, which is based on the I-topology T· of FNSs, under this framework, we will consider further later. References [1] [2] [3] [4] [5]

C.L. Chang, Fuzzy topological spaces, J. Math. Anal. Appl. 24 (1968) 182–190. N.F. Das, P. Das, Fuzzy topology generated by fuzzy norm, Fuzzy Sets and Systems 107 (1999) 349–354. D. Dubois, H. Prade, Fuzzy elements in a fuzzy set, Proc. 10th Internat. Fuzzy Systems Assoc. (IFSA) Congress, Beijing, 2005, pp. 55–60. J.-x. Fang, On local bases of fuzzy topological vector spaces, Fuzzy Sets and Systems 87 (1997) 341–347. J.-x. Fang, G.-a. Song, -contractor and the solutions for nonlinear operator equations in fuzzy normed spaces, Fuzzy Sets and Systems 121 (2001) 267–273. [6] C. Felbin, Finite dimensional fuzzy normed linear space, Fuzzy Sets and Systems 48 (1992) 239–248. [7] T.E. Gantner, R.C. Steinlage, R.H. Warren, Compactness in fuzzy topological spaces, J. Math. Anal. Appl. 62 (1978) 547–562. [8] in: U. Höhle, S.E. Rodabaugh (Eds.), Mathematics of Fuzzy Sets: Logic, Topology, and Measure Theory, The Handbooks of Fuzzy Sets Series, vol. 3, Kluwer Academic Publishers, Dordrecht, 1999. [9] O. Kaleva, S. Seikkala, On fuzzy metric spaces, Fuzzy Sets and Systems 12 (1984) 215–229. [10] A.K. Katsaras, Fuzzy topological vector spaces I, Fuzzy Sets and Systems 6 (1981) 85–95. [11] Y.-m. Liu, M.-k. Lao, Fuzzy Topology, World Scientific, Singapore, 1997. [12] R. Lowen, Fuzzy topological spaces and fuzzy compactness, J. Math. Anal. Appl. 56 (1976) 621–633.

2750

J.-x. Fang / Fuzzy Sets and Systems 157 (2006) 2739 – 2750

[13] C.-x. Wu, J.-x. Fang, QL-type fuzzy topological vector space, Chinese Ann. Math. 6A (3) (1985) 355–364. [14] j.-z. Xiao, X.-h. Zhu, On linearly topological structure and property of fuzzy normed linear space, Fuzzy Sets and Systems 125 (2002) 153–161. [15] j.-z. Xiao, X.-h. Zhu, Fuzzy normed space of operators and its completeness, Fuzzy Sets and Systems 133 (2003) 389–399. [16] j.-z. Xiao, X.-h. Zhu, The uniform boundedness principles for fuzzy normed spaces, J. Fuzzy Math. 12 (2004) 543–560.