FUZZY ACTS OVER FUZZY SEMIGROUPS AND SHEAVES

Transcription

1 Iranian Journal of Fuzzy Systems Vol. 11, No. 4, (2014) pp FUZZY ACTS OVER FUZZY SEMIGROUPS AND SHEAVES M. HADDADI Abstract. Although fuzzy set theory and sheaf theory have been developed and studied independently, Ulrich Hohle shows that a large part of fuzzy set theory is in fact a subfield of sheaf theory. Many authors have studied mathematical structures, in particular, algebraic structures, in both categories of these generalized (multi)sets. Using Hohle s idea, we show that for a (universal) algebra A, the set of fuzzy algebras over A and the set of subalgebras of the constant sheaf of algebras over A are order isomorphic. Then, among other things, we study the category of fuzzy acts over a fuzzy semigroup, so to say, with its universal algebraic as well as classic algebraic definitions. 1. Introduction and Preliminaries Although fuzzy set theory and sheaf theory have been developed and studied independently, Hohle [17, 18] gives the close relation between them. Both are in some sense multisets, and hence are generalizations of ordinary (single) sets. This property makes both theories strongly useful in mathematics and many of its applications. Algebraic structures have specially been studied in both disciplines, see, for example, [3, 4, 5, 20, 8, 9, 13, 11, 10]. Using Hohle s idea, we show that for a (universal) algebra A, the poset of fuzzy algebras over A and the poset of subalgebras of the constant sheaf over A are order isomorphic. Thus, one can use the power and the tools in both theories to study mathematics in these fields. The main purpose of this paper is to give briefly the above mentioned relation between fuzzy algebras and sheaves of algebras (Section 2) to open another door to study fuzzy algebras, and to study the category of fuzzy (so to say, universal algebraic) acts over a crisp semigroup and fuzzy (so to say, classic algebraic) acts over a fuzzy semigroup (Sections 3 and 4). Let us recall the following definitions needed in the sequel, and introduce some notations used in this paper. Definition 1.1. A set X together with a function µ : X [0, 1] is called a fuzzy set (over X) and is denoted by (X, µ) or X (µ). We call X the underlying set and µ the membership function of the fuzzy set X (µ), and µ(x) [0, 1] is the grade of membership of x in X (µ). Received: October 2011; Revised: May 2014; Accepted: June 2014 Key words and phrases: Fuzzy algebra, Fuzzy act, Sheaf.

2 62 M. Haddadi If µ is a constant function with value a [0, 1], X (µ) is denoted by X (a). The fuzzy set X (1) is called a crisp set and may sometimes simply be denoted by X. For a fuzzy set X (µ) and α [0, 1], X α (µ) = X (µ) [α := {x X µ(x) α} is called the (closed) α-cut or the (closed) α-level set of the fuzzy set X (µ), and := {x X µ(x) > α} is called the open α-cut or the open α-level set. The X (µ) (α cuts S(µ) = X (µ) (0 = {x µ(x) > 0} and C(µ) = X (µ) 1 = {x µ(x) = 1} are called the support and the core (also called, the kernel, the singular cut) of µ, respectively. A fuzzy function from X (µ) to Y (η), written as f : X (µ) Y (η), is an ordinary function f : X Y such that the following is a fuzzy triangle: X µ [0, 1] f Y η meaning that µ ηf (that is, µ(x) ηf(x) for all x X). The set of all fuzzy sets with a fixed underlying set X is called the fuzzy power or the set of fuzzy subsets of X and is denoted by FSubX. Clearly fuzzy sets together with fuzzy functions between them form a category denoted by FSet. Note that the category FSet is concrete over Set. Also, the left adjoint (free functor) F to the obvious forgetful functor U : FSet Set takes a set X to the fuzzy set X (0), and a function f : A B to itself. Since the category FSet has the free object over a singleton set, a fuzzy function f : X (µ) Y (η) is monic if and only if f : X Y is one to one and so, in this case, one may write X (µ) Y (η). Also, if f = id X : X X, then X (µ) X (η) if µ η, that is, µ(x) η(x), for all x X. In the case where X Y and f is the inclusion function, X (µ) is called a fuzzy subset of Y (η), written as X (µ) Y (η). Note that one can easily see that the partially ordered set FSubX, with the above order, is a complete lattice with the lattice operations of membership functions as ( µ i )(x) := µ i (x) ( µ i )(x) := µ i (x) Also, it easily follows that for all X (µ), X (ν) FSubX X (µ) X (ν), [0, 1] X (µ) X (ν). γ,, γ [0, 1] X γ (µ) X (µ) X (µ) = X (ν) X (µ) = X (ν),, for all [0, 1]. 2. Universal Algebra in the Categories Sh[0,1] and FSet In this section, we first recall the general notion of a (universal) algebra in an ordinary category and then, using Hohle s idea [17, 18], we give the relation between

3 Fuzzy Acts over Fuzzy Semigroups and Sheaves 63 the categories of algebras in the categories FSet of fuzzy sets and Sh[0,1] of sheaves on the frame (complete Heyting algebra) [0,1]. Definition 2.1. Let E be a finitely complete category (or at least with finite products). Then a (universal) algebra in E is an entity (A, ( A ) Λ ), where A is an object of E, Λ is a set and for each Λ, the th operation A : A n A is a morphism in E where each n is a finite cardinal number and A n is the n th power of A. The family τ = (n ) Λ is called the type of this algebra. The algebra (A, ( A ) Λ ) is simply denoted by A. Also, a homomorphism h : A B from an algebra (A, ( A ) Λ ) to an algebra (B, ( B ) Λ ), both of type τ in E, is a morphism in E such that for each Λ, B h n = h A. Algebras (of type τ) in E and homomorphisms between them form a category denoted by AlgE. If E is the category Set of sets, AlgE is simply denoted by Alg (see [6]). Usually for the classical algebraic structures, such as groups, rings, and vector spaces, the operations satisfy some identities (such as associativity, commutativity, etc.). The full subcategory of AlgE whose objects are all algebras in E satisfying a set Σ of equations is denoted by Mod(Σ, E). No need to mention the importance of sheaves over topological spaces or frames and complete Heyting algebras, (that is, complete lattices in which meets distribute over arbitrary joins), see, for example, [22, 23]. Let us very briefly recall this notion and sheaves of algebras. First note that every partially ordered set P can be considered as a category P whose objects are the elements of P and for every a, b P, { { } if a b Hom(a, b) = otherwise Thus, in particular, every frame is a category. Now, recall that a presheaf (of sets) on a frame L is a functor F : L op Set in which for each β α, the corresponding function F α F β is denoted by ρ α β or β and called a restriction. That is, for γ β α, ρ α α = id F α & ρ α β ρ β γ = ρ α γ. Also, morphisms between presheaves are just natural transformations η : F G; that is, η is an L indexed family (η α ) α L of maps F α ηα Gα such that for each pair β α, the following diagram is commutative: F α ηα Gα ρ α β F β ηβ Gβ A presheaf F on L is called a sheaf if the following two conditions are satisfied: F is a mono presheaf (or separated), that is, if α = i I α i and s, t F α, such that s α i = t α i, for all i I, then s = t. ρ α β

4 64 M. Haddadi F satisfies the patching (or gluing) condition, that is, if α = i I α i and (s i ) i I i I F α i with s i α i α j = s j α i α j, for all i, j I, then there exists s F α (unique by separatedness) with s α i = s i, for all i I. In the following, the frame L is considered to be the unit interval [0, 1]. So every presheaf F on [0, 1] is an indexed family (F α) α [0,1] with the restriction functions (ρ α β : F α F β) α β. Specially, for an arbitrary set A, the corresponding constant sheaf à is defined by Ãα = A, if α 0, and Ã0 = { }. One can easily see that a subsheaf G of a sheaf F is a sheaf with Gα F α, for each α L, and each ρ α β : Gα Gβ is the restriction of each ρα β : F α F β. The following lemma is a characterization of the subsheaves of a constant sheaf on L = [0, 1]. Lemma 2.2. F is a subsheaf of a constant sheaf on [0, 1] if and only if (F α) α [0,1] fulfills any one of the following equivalent conditions: (i) F 0 = { }, F α F β, where 0 < β α. (ii) 0<β<α F β = F α. Proof. It is clear that every subsheaf of a constant sheaf à is of the form (i). So we just show that (i) and (ii) are equivalent. Let a sheaf F satisfy condition (i). Thus, F α F β, for every 0 < β α. Hence F α 0<β α F β. On the other hand, let a 0<β<α F β. Then consider a β = a F β, for every β < α, with the property that ρ β β (a β ) = a β, for every β β < α. Since α = β<α β, (by the patching property of sheaves) there is a α F α such that ρ α β (a α) = a β = a. That is, a F α and hence 0<β<α F β = F α. Now let F be a sheaf satisfying (ii). Then F, being a sheaf, gives that F 0 = { } and for 0 < β α, F α F β, since 0<β<α F β = F α. It is easy to see that finite limits in the category ShL of sheaves on L are computed componentwise. In particular, finite product of sheaves F 1,..., F n, is computed as ( n i=1 F i)α = n i=1 F iα, for every α L. Since ShL is finitely complete (in fact a topos), algebras of type τ can be defined in ShL as in Definition 2.1. That is to say, a sheaf of algebras of type τ on L is a sheaf A ShL if its n-ary operations is a natural transformation A : A n A. It can be proved that A is an algebra of type τ in ShL if and only if each Aα is an ordinary algebra of the same type. Also A satisfies a set Σ of equations if and only if each Aα satisfies Σ. Now by Lemma 2.2 we have the following. Corollary 2.3. For every A Alg, the corresponding constant sheaf à is an algebra of the same type in ShL. Also, every subalgebra of à is represented by the chain (F (α)) α (0,1] of subalgebras of A, where F (α) F (β) for 0 < β α, and F (0) = { }. The set of all subalgebras of the constant sheaf à (of algebras) is denoted by ShSub a Ã.

5 Fuzzy Acts over Fuzzy Semigroups and Sheaves 65 Finally in this section, we introduce the category of (universal) algebras of type τ in the category FSet of fuzzy sets. But first we note that the category FSet is finitely complete, in particular the product of fuzzy subsets A (µ) and B (η) is (A B) (µ η), and thus one can define (universal) algebras in this category as follows. Definition 2.4. (i) By an n-array fuzzy operation on a fuzzy set A (µ) in the category FSet we mean a fuzzy function from A n to A. More explicitly, an n-array operation on a fuzzy set A (µ) gives the following fuzzy triangle. n A n i=1 µ [0, 1] meaning that for every a 1,..., a n A, µ(a i ) µ( A (a 1,..., a n )). i=1,...,n A µ (ii) A (universal) algebra of type τ in the category FSet, called a fuzzy (universal) algebra, is a fuzzy set A (µ) together with a family { i : A n A} i I of fuzzy operations. (iii) A fuzzy homomorphism f : A (µ) B (η) between fuzzy algebras of the same type is a fuzzy map as well as an algebra homomorphism (preserving the operations) from A to B. The set of all fuzzy algebras over a fixed algebra A is denoted by FSub a A, and the category of all fuzzy algebras of type τ and the fuzzy homomorphisms between them is denoted by FAlg(τ), or simply by FAlg. Lemma 2.5. Let A (µ) FAlg. Then every α-cut A (µ) α is an (ordinary) algebra. Proof. Suppose that a 1,..., a n belong to A α (µ). Then we have α i=1,...,n µ(a i) µ( A (a 1,..., a n )), which proves the result. Note that, for an ordinary (universal) algebra A Alg, the sets ShSub a à and FSub a A are partially ordered, with the obvious orders. In the next theorem we show that they are order isomorphic. Theorem 2.6. Let A Alg. ShSub a à are order isomorphic. Then, the partially ordered sets FSub a A and Proof. Define the map G : FSub a A ShSub a à by G(A (µ) ) := F µ in which F µ (α) := A (µ) α, if α 0, and F µ (0) := { }. Corollary 2.3 and Lemma 2.5 ensure that F µ is a subalgebra of the constant sheaf Ã. Note that G is order-preserving, because if A (ν) A (µ) then, for every [0, 1], A (ν) A(µ). Hence F ν F µ. Now we define H : ShSub a à Fsub a A by H(F ) := A (µ F ) in which µ F (a) := a F (), for every a A. We note that H is order-preserving. Because, if F F, then F (α) F (α) for every α [0, 1], and hence a F () a F ().

6 66 M. Haddadi Also A (µ F ) is a fuzzy algebra of type τ, because for every n-ary operation A, and a 1,..., a n A, if µ F (a i ) 0, for every i = 1,..., n, then a j F ( n i=1 µ F (a i )), for every j = 1,..., n. So A (a 1,..., a n ) F ( n i=1 µ F (a i )) and hence n i=1 µ F (a i ) µ F ( B (a 1,..., a n )). Also if there exists 1 j n such that µ F (a j ) = 0, then obviously, 0 = n i=1 µ F (a i ) µ F ( B (a 1,..., a n )). Now we show that H and G are inverses of each other. For each fuzzy algebra A (µ), we have: µ Fµ (a) = a F µ() = µ(a) = µ(a) and also F µf (α) = {a A µ F (a) α} = {a A α} = a F () { F (α) if α 0 { } otherwise (1) The last equality is because F (α) {a A a F () α}, for α is one of,s in (1), and since a F () α, there exists 0 with a F ( 0 ) and α 0. Hence for F ( 0 ) F (α), a F (α). 3. Fuzzy Acts over Fuzzy Semigroups Universal algebras having a semigroup (monoid or group) S of unary operations have always been of interest to mathematicians, specially to computer scientists and logicians. The algebraic structures so obtained are called S-acts (or S-sets, S-automata, S-operands, S-polygons, S-systems, see [12, 19, 2, 7]). As a very interesting example, used in computer sciences as a convenient means of algebraic specification of process algebras (see [14, 15]), consider the monoid (N, ) acting on sets, where N is the set of natural numbers and N = N { } with n <, n N and m n = min{m, n} for m, n N (see, for example, [14, 16]). In this section we are going to define the notion of the actions of a (fuzzy) semigroup on a fuzzy set, which will be studied in this and the last section of this paper. First note that, for a semigroup S and s S, we get the left and the right translations s, ρ s : S S defined by s (x) = sx, ρ s (x) = xs. Now, according to the definition of a fuzzy (universal) algebra given in Section 2, we have the following definition of a fuzzy semigroup.

7 Fuzzy Acts over Fuzzy Semigroups and Sheaves 67 Definition 3.1. A semigroup S together with a function ν : S [0, 1] is called a fuzzy semigroup if its multiplication is a fuzzy function: for every s, r S, ν(s) ν(r) ν(sr); that is, the following is a fuzzy triangle: S S ν ν [0, 1] S S in which S : S S S is the multiplicatin of the semigroup S. If S has an identity 1, one usually add the condition ν(1) = 1. Now, note that, for a (crisp) semigroup S, a (crisp) set A can be made into an (ordinary) S-act in the following two equivalent ways: (Universal algebraic) The set A together with a family ( s : A A) s S of unary operations satisfying (st)a = s(ta) (and 1a = a, if S has an identity) where sa = s (a). (Classic algebraic) The set A together with a function : S A A satisfying (st)a = s(ta) (and 1a = a, if S has an identity) where sa = (s, a). Now, having these two, so called, universal and classic actions of S on A, we get the following two, not necessarily equivalent, definitions for a fuzzy act over a fuzzy monoid. Definition 3.2. Let S (ν) be a fuzzy semigroup and A (µ) be a fuzzy set such that A is an S-act, as defined above. Then, A (µ) is called: (Universal) A fuzzy S-act (or fuzzy S (1) -act, where S (1) is the crisp semigroup S to emphasize fuzziness) if each s is a fuzzy function; that is µ(a) µ(sa), for every s S and a A (with no mention of ν). That is, for every s S, the following triangle is fuzzy: ν A µ [0, 1] s A µ (Classic) A fuzzy S (ν) -act if : S A A is a fuzzy function; that is, ν(s) µ(a) µ(sa), for every s S and a A. That is, the following triangle is fuzzy: S A S ν µ µ [0, 1] Note that Universal implies Classic, and if S (1) = S is a (crisp) semigroup, then ν(s) µ(a) = 1 µ(a) = µ(a), and so the above two definitions are equivalent.

8 68 M. Haddadi Also, every fuzzy semigroup S (ν) is naturally a fuzzy S (ν) -act (classically) and S (1) = S is a fuzzy S (1) = S-act (universally, and hence classically). Also, if S (ν) is a fuzzy left ideal, then it is a fuzzy S-act (universally, and hence classically). A morphism between fuzzy S (ν) -acts (universal or classic), also called an S (ν) - map is simply an S-map (that is, a map together with action-preserving property) as well as a fuzzy function. The set of all (fuzzy) S (ν) -acts with a fixed A is denoted by S (ν) -FSubA, and the category of all fuzzy S (ν) -acts with morphisms between them is denoted by S (ν) -FAct. Also note that, since an S-act A is naturally a (unary) universal algebra, the universal definition of fuzzy acts, being compatible with Definition 2.4, may be considered to be more algebraic than the classic one. Thus, by Theorem 2.6, the partially ordered set S (1) -FSubA of (universal) fuzzy S-acts is order isomorphic to the partially ordered set ShSub a à of subsheaves of the constant sheaf of S-acts Ã. In particular, we have the counterpart of Lemma 2.5 as follows: Theorem 3.3. An S-act A with µ : A [0, 1] is a fuzzy S = S (1) -act if and only if for every α [0, 1], A (µ) α is an ordinary S-subact of A. Proof. One way is clear by Lemma 2.5. Now let for every α [0, 1], A (µ) α be an ordinary subact of A, s S, and a A. Then, since a A (µ) µ(a) and A(µ) µ(a) is a subact of A, sa A µ µ(a). That is µ(a) µ(sa), for every s S and a A. As we have seen, Lemma 2.5 and Theorem 2.6 are true for the universal definition of fuzzy acts. In the following we are going to give the counterparts of these results for the classic definition of fuzzy acts. First recall from [11] that, for a semigroup S in Sh[0, 1] (sheaf of semigroups), an S-sheaf is a sheaf A together with a natural transformation η : S A A such that (Aα, η α ) is an Sα-act, for each α [0, 1] and ρ α β : Aα Aβ is equivariant over ρ α β : Sα Sβ for every β α; that is, ρα β (sa) = ρ α β (s)ρα β (a) (sa β = (s β)(a β)), for every s Sα, a Aα. Since, in view of Theorem 2.6, every fuzzy semigroup S (ν) corresponds to a sheaf F ν of semigroups, we can consider F ν -sheaves. Now we claim that the partially ordered set S (ν) -FSubA of fuzzy S (ν) -subacts of A is order isomorphic to the partially ordered set F ν -ShSubà of subalgebra of the constant F ν-sheaf à of algebras. To prove the claim, first we note the following lemma. Lemma 3.4. Let A be an S-act. Then, for every fuzzy semigroup S (ν), the constant sheaf à is an F ν -sheaf. Proof. Since A is an S-act, we can define the natural transformation η : F ν à à to be the family (η α ) α [0,1] in which η α maps every (s, a) (F ν A)α = F ν α Aα to sa, for every α [0, 1]. We note that η is a natural transformation because if β α, then ρ α β : (F ν Ã)α (F ν Ã)β is nothing but the product of the inclusion map

9 S (ν) α S (ν) β Fuzzy Acts over Fuzzy Semigroups and Sheaves 69 and id : Aα Aβ. So we have the following commutative diagram (s, a) sa ρ α β (s, a) sa ρ α β Also, in view of the above diagram, it is clear that ρ α β : Aα Aβ is equivariant over ρ α β : S(ν) α S (ν) β for every β α. The following is the counterpart of Theorem 3.3 which leads us to our claim. Theorem 3.5. A (µ) is a fuzzy S (ν) -act if and only if A (µ) [0, 1]. That is F µ is an F ν -sheaf. is an S (ν) -act for every Proof. First suppose that A (µ) is a fuzzy S (ν) -act of A, [0, 1], s S (ν), and a A (µ). Then since ν(s) µ(a) µ(sa), sa A(µ). Now let S(ν) be A (ν) -act, for every [0, 1]. Then for every a A and s S, take = ν(s) µ(a). Since a A (µ) and s S (ν), sa A(µ). Hence ν(s) µ(a) µ(sa). Theorem 3.6. The partially ordered set S (ν) -FSubA is order isomorphic to the partially ordered set of subsheaf acts over [0, 1] of the constant F ν -sheaf à which is denoted by F ν -ShSubÃ. Proof. Define the order preservig map G : S (ν) -FSubA F ν -ShSubà to be G(A (µ) ) := F µ. We have already seen, in Theorems 2.6 and Lemma 3.4, that F µ is a subact of the constant sheaf à and à is an F ν-sheaf and in Theorem 3.5 we saw that F µ is an F ν -sheaf. It is easily seen that if A (µ) A (γ), then F µ F γ. Also define the functor H : F ν -ShSubà S(ν) -FSubA to be H(F ) := µ F (as we defined in Theorem 2.6). Now we just show that µ F is a fuzzy S (ν) -subact of A. So consider µ F (sa), if sa F α for some α 0, then µ F (sa) = sa F. Specially for 0 = ν(s) µ F (a), sa F 0. Hence ν(s) µ F (a) µ F (sa). Also if sa / F α, for every α [0, 1], then a / F α or s / F α, for every α [0, 1], because otherwise, if a F α and s F β, then sa F (α β). So if µ F (sa) = 0 then µ F (a) ν(s) = 0. Now in view of Theorem 2.6 there is nothing to prove. 4. The Category of Fuzzy Acts In this section we give some categorical properties of fuzzy S (ν) -acts (see [1, 21] for categorical notions.) Theorem 4.1. The product of two fuzzy S (ν) -acts A (µ) and B (η) in the category S (ν) -FAct is (A B) (µ η). Proof. Since (µ η)(a, b) ν(s) = µ(a) η(b) ν(s) ν(s) µ(sa) η(sb) = (µ η)(s(a, b)), for every (a, b) A B and s S, (A B) (µ η) is an S (ν) -act in which the action of S over A B is defined component wise, that is (s(a, b)) = (sa, sb).

10 70 M. Haddadi Also, since the following triangles are clearly fuzzy triangles, the projections are fuzzy maps: π B π A B A B A η [0, 1] Now let C (γ) be an S (ν) -act with fuzzy S (ν) -maps f and g in the following diagram of fuzzy triangles: B η f C µ η γ [0, 1] Then, since γ µg and γ ηf, γ µg ηf, the S (ν) -map (g, f) : C A B with (g, f)(c) = (g(c), f(c)) is a fuzzy map, that is, makes the diagram (f,g) C A B γ µ η g µ µ [0, 1] a fuzzy triangle. It is easy to verify that (g, f) : C A B is the unique fuzzy S (ν) - map which can commute the above diagram. This clearly proves the universality of (A B) (µ η), and hence we get the result. Also, note that the terminal object in S (ν) -FSet, which is the product of the empty family, is T : { } [0, 1] which maps to 1 with the constant action. The following is easily proved. Theorem 4.2. The coproduct of two fuzzy S (ν) -acts A (µ) and B (η) is the fuzzy S-act (A B) (µ η), where µ ν A = µ and µ η B = η. Theorem 4.3. The equalizer of two fuzzy S (ν) -act maps f, g : A (µ) B (η) is the fuzzy S (ν) -act E (e), where E = {a A f(a) = g(a)} and e := µ E : E [0, 1]. Proof. First note that E (e) is clearly a fuzzy S (ν) -subact of A (µ) equalizing f and g. Now suppose C (γ) is a fuzzy S (ν) -act and h : C (γ) A (µ) is a fuzzy S (ν) -map (γ µh) such that fh = gh. Then, one can easily see that the map h : C (γ) E (e), taking every c C to h(c) E, is the unique fuzzy S (ν) -map which gives the universality of E (e), and hence we get the result. Theorem 4.4. The pullback of two fuzzy S (ν) -maps f : A (µ) C (γ) and g : B (η) C (γ) is the fuzzy equalizer of the two fuzzy S (ν) -maps f π A, g π B : (A B) (µ η) C (γ) Now using the above theorem about pullbacks, we easily get the following facts. A

11 Fuzzy Acts over Fuzzy Semigroups and Sheaves 71 Corollary 4.5. The inverse image of B (η) B (γ) along a fuzzy S (ν) -map f : A (µ) B (γ) is f 1 (η) : E [0, 1] with E = {(a, b) A B f(a) = b} and f 1 (η)(a, b) = µ(a) η(b). So we can consider f 1 (η) : A [0, 1] to be f 1 (η)(a) = µ(a) η(f(a)) which is a fuzzy S (ν) -subact of µ. Because f 1 (η)(a) = µ(a) η(f(a)) µ(a) γ(f(a)) = µ(a), for every a A. In the rest of this section we are going to construct a sort of image and inverse image of an S (ν) -map between fuzzy S (ν) -acts in the category S (ν) -FSubA. To do this, first we give the following definition. Definition 4.6. Let A (µ) be a fuzzy subact of A and B be a set, and f : A B be a map. Then we define the fuzzy subset f(µ) : B [0, 1] to be: f(µ)(b) := {µ(a) f(a) = b}, for every b B. Theorem 4.7. Suppose that A (µ) is a fuzzy S (ν) -subact of A, B is an S-act, and f : A B is an S-map, then f(µ) is a fuzzy S (ν) - subact of B. Proof. To prove, consider f(µ)(sb) = {µ(a) f(a) = sb} and f(µ)(b) := {µ(a) f(a) = b}. But if f(a) = b, then sf(a) = f(sa) = sb. Now since ν(s) µ(a) µ(sa), ν(s) {µ(a) f(a) = b} {µ(a) f(a) = sb} and hence f(µ)(b) f(µ)(sb). Theorem 4.8. Let f : A B be an S-map between S-acts A and B, and B (µ) be a fuzzy S (ν) -subact of B. Then µ(f) : A [0, 1], defined by µ(f)(a) := µ(f(a)), is a fuzzy S (ν) -subact of A. Proof. ν(s) µ(f)(a) = ν(s) µ(f(a)) µ(sf(a)) = µ(f(sa)) = µ(f)(sa), for every a A and s S. In view of Theorems 4.7 and 4.8 we can consider the functors f 1 : S (ν) - FSubB S (ν) -FSubA to be f 1 (B (µ) ) = A (µ(f)) and call it the inverse image of f, and f : S (ν) -FSubA S (ν) -FSubB to be f(a (η) ) = B (f(µ)) and call it the image of f, where f : A B is an S-map between S-acts A and B. The terms used for the above functors are justified, since f is a left adjoint of f 1. More explicitly: Proposition 4.9. If f : A B is an S-map between S-acts A and B, and B (µ) and A (ν) are respectively fuzzy S (ν) -subacts of B and A, then f(µ(f)) µ and η (f(η))(f). Proof. For every b B we have: f(µ(f))(b) = {µ(f)(a) f(a) = b} = {µ(f(a)) f(a) = b} { µ(b) if f = 1 (b) 0 other wise µ(b)

12 72 M. Haddadi Also we have: ((f(η))(f))(a) = (f(η))(f(a)) = {η(x) f(x) = f(a)} η(a) for every a A. So η (f(η))(f). Acknowledgments. I am grateful to professor Ebrahimi for many conversations on this and related matters, and for his enthusiasm for my project and gratefully acknowledge the referees for carefully reading the article and making helpful comments. References [1] J. Adámek, H. Herrlich and G. E. Strecker, Abstract and concrete categories, John Wiley and Sons, [2] B. Banaschewski, Equational Compactness of G-sets, Canadian Mathematical Bulletin, 17(1) (1974), [3] R. Belohlavek, Birkhoff variety theorem and fuzzy logic, Archive for Mathematical Logic, 42 (2003), [4] R. Belohlavek and V. Vychodil, Algebras with fuzzy equalities, Fuzzy Sets and Systems, 157 (2006), [5] I. Bosnjak, R.Madarasz and G.Vojvodic, Algebra of fuzzy sets, Fuzzy Sets and Systems, 160 (2009), [6] S. Burris and H. P. Sankapanavar, A course in universal algebra, Springer-Verlag, [7] S. Burris and M. Valeriote, Expanding varieties by monoids of endomorphisms, Algebra Universalis, 17(2) (1983), [8] M. M. Ebrahimi, Algebra in a Grothendieck topos: injectivity in quasi-equational classes, Pure and Applied Algebra, 26(3) (1982), [9] M. M. Ebrahimi, Equational compactness of sheaves of algebras on a notherian local, Algebra Universalis, 16 (1983), [10] M. M. Ebrahimi and M. Haddadi, Essential pure monomorphisms of sheaves of group actions, Semigroup Forum, 80 (2010), [11] M. M. Ebrahimi, M. Haddadi and M. Mahmoudi, Equational compactness of G-sheaves, Communucations in Algebra, 40 (2012), [12] M. M. Ebrahimi and M. Mahmoudi, The category of M-sets, Italian Journal of Pure and Applied Mathematics, 9 (2001), [13] M. M. Ebrahimi and M. Mahmoudi, Purity of G-sheaves, Submitted, [14] H. Ehrig, F. Parisi-Presicce, P. Boehm, C. Rieckhoff, C. Dimitrovici and M. Grosse-Rhode, Algebraic data type and process specifications based on projection Spaces, Lecture Notes in Computer Science, 332 (1988), [15] H. Ehrig, F. Parisi-Presicce, P. Boehm, C. Rieckhoff, C. Dimitrovici and M. Grosse-Rhode, Combining data type and recursive process specifications using projection algebras, Theoretical Computer Science, 71 (1990), [16] H. Herrlich and H. Ehrig, The construct PRO of projection spaces: its internal structure, Lecture Notes in Computer Science, 393 (1988), [17] U. Hohle, Fuzzy sets and sheaves. Part I Basic concepts, Fuzzy Sets and Systems, 158 (2007), [18] U. Hohle, Fuzzy sets and sheaves. Part II: Sheaf-theoretic foundations of fuzzy set theory with applications to algebra and topology, Fuzzy Sets and Systems, 158 (2007), [19] M. Kilp and U. Knauer and A. Mikhalev, Monoids, Acts and Categories, New York, 2000.

13 Fuzzy Acts over Fuzzy Semigroups and Sheaves 73 [20] T. Kuraoka, Formulas on the lattice of fuzzy subalgebras in universal algebra, Fuzzy Sets and Systems, 158 (2007), [21] S. Maclane, Categories for the working Mathematicians, Springer Verlag, [22] S. Maclane and I. Moerdijk, Sheaves in Geometry and Logic, Springer Verlag, [23] R. Tennison, Sheaf Theory, Cambridge University Press, M. Haddadi, Department of Mathematics, Statistics and Computer Sciences, Semnan University, Semnan, Iran address: haddadi