THE UNIVERSAL SEMILATTICE COMPACTIFICATION OF A SEMIGROUP

The universal abelian, band, and semilattice compactifications of a semitopological semigroup are characterized in terms of three function algebras. Some relationships among these function algebras and some well-known ones, from the universal compactification point of view, are also discussed.


Introduction.
The notion of semigroup compactification has been produced in several principal ways, in whose main approach the Gelfand-Naimark theory of commutative C * -algebras is employed.In fact, the spectrum of every m-admissible algebra of functions is a semigroup compactification.Moreover, some of these compactifications enjoy a universal property P .For instance, De Leeuw and Glicksberg in their influential paper [2], characterized the universal property of (weakly) almost periodic compactification.The existence of the universal P -compactification (using the subdirect product methods) for a broad variety of properties P , is guaranteed by Junghenn and Pandian [7].The construction of some of the better known universal P -compactifications in terms of m-admissible algebras of functions are collected in Berglund et al. [1], which is our ground reference.The universal right simple, left simple, and group compactifications are characterized in terms of some types of distal functions [6].In two recent papers [9,10], Pandian has examined the universal mapping property of generalized distal, and quasiminimal distal functions.Also, in an earlier paper [3], we have characterized the universal nilpotent group compactification.The present paper deals with the construction of three m-admissible algebras AB, BD, and SL, which characterize the universal abelian, band, and semilattice compactifications of a semitopological semigroup.

Preliminaries.
For background and notations we follow Berglund et al. [1] as much as possible.In what follows, S is a semitopological semigroup unless otherwise stipulated.A (semigroup) compactification of S is a pair (ψ, X), where X is compact, Hausdorff, right topological semigroup and ψ : S → X is a continuous homomorphism with dense image such that, for all s ∈ S, the mapping x → ψ(s)x : X → X is continuous.
The C * -algebra of all continuous bounded complex-valued functions on a topological space Y is denoted by C(Y ).For C(S) left and right translations, L s and R t , are defined for all s, t ∈ S by is in F for all f ∈ F and µ ∈ S F ( = the spectrum of F ).If so, S F under the multiplication µν = µ • T ν (µ, ν ∈ S F ), furnished with the Gelfand topology, makes (ε, S F ) a compactification (called the F -compactification) of S, where ε : S → S F is the evaluation mapping.Conversely, if (ψ, X) is a compactification of S, then ψ * (C(X)) is an m-admissible subalgebra of C(S), where ψ * is the dual mapping of ψ, and this correspondence between compactifications of S and m-admissible subalgebras of C(S) is one-to-one (see [1,Thm. 3.1.7]).
A compactification (ψ, X) of S, possessing a certain property P , is called the universal P -compactification if for any other compactification (ϕ, Z), having the property P , there exists a homomorphism π : Some of the usual m-admissible subalgebras of C(S), that are needed in the sequel, are the left multiplicatively continuous, weakly almost periodic, almost periodic, strongly almost periodic, distal, minimal distal, and strongly distal functions on S.These are denoted by LMC, WAP , AP , SAP ,D,MD and SD, respectively.We also write GP for MD ∩ SD, LZ for {f ∈ C(S) : f (st) = f (s) for all s, t ∈ S}, and RZ for {f ∈ C(S) : f (st) = f (t) for all s, t ∈ S}.Here, and also for other emerging spaces, when there is no risk of confusion, we have suppressed the letter S from the notation.For ease of reference, we mention the next proposition which describes the universal mapping properties of these m-admissible algebras.
Proposition 2.1.See [1,Chap. 4] and [6,Thm. 3.4].The LMC, WAP , AP , SAP ,D,MD, SD, GP , LZ , and RZ -compactifications are universal with respect to the properties of being a (right topological) semigroup, a semitoplogical semigroup, a topological semigroup, a topological group, an inflation of a rectangular group, a left simple semigroup, a right simple semigroup, a group, a left zero semigroup, and a right zero semigroup, respectively.

The main results.
To follow the main objective, we examine the properties of AB and BD, where and for all nets {s α } and {t α } in S for which the relevant pointwise limits exist.
Also, we write SL for AB ∩ BD.The next lemma, which requires a routine proof, characterizes AB and BD in terms of the elements of S WAP and S LMC , respectively.

Lemma 3.1. (i) A function f ∈ WAP is in AB if and only if µν(f ) = νµ(f ) and
The following theorem states the main properties of AB, BD, and SL.Theorem 3.2.AB, BD, and SL are those m-admissible subalgebras of C(S), whose corresponding compactifications of S are universal with respect to the properties of being an abelian semigroup, a band, and a semilattice, respectively.
Proof.It is enough to prove the conclusion for AB and BD.Using Lemma 3.1, the m-admissibility of AB and BD can be easily demonstrated, and also it follows that S AB and S BD are abelian and a band, respectively.Let (ψ, X) be an abelian compactification of S, then C(X) = AB(X) and so ψ * (C(X)) = ψ * (AB(X)) ⊆ AB(S), where the latter inclusion can be easily verified.Thus, (ε, S AB ) is the universal abelian compactification of S. Similarly, to see that (ε, S BD ) is universal with respect to the property of being a band, it is enough to show that for any other band compactification (ϕ, Z) of S, ϕ * (C(Z)) ⊆ BD(S).For this, let π : (ε, S LMC ) → (ϕ, Z) be the canonical homomorphism whose existence is guaranteed by the universal property of (ε, S LMC ).
It is trivial that BD ⊆ BD c (with the equality holding in the compact case), where and f st 2 u = f (stu) for all s, t, u ∈ S .(3.3)The joint continuity of the multiplication of S AP implies that BD ∩ AP = BD c ∩ AP .Furthermore, S SL is a compact semitopological semilattice, so by Lawson's (joint continuity) theorem [8], SL ⊆ AP .Thus, SL = AP ∩ BD c ∩ AB; more precisely: The universal properties of (ε, S BD ) and (ε, S D ) imply that (ε, S BD ∩D ) is universal with respect to the property of being a rectangular band [1,Exercise 1.1.48].Furthermore, since every such rectangular band is a topological semigroup, BD ∩D ⊆ AP which implies that BD ∩D = BD c ∩D ∩ AP .On the other hand, an adaptation of Junghenn's ideas in the proof of Proposition 3.10 of [6] implies that BD ∩D = LZ ∪ RZ = LZ ⊗ RZ , where LZ ∪ RZ is the C * -subalgebra of C(S) generated by LZ ∪ RZ and LZ ⊗ RZ is the topological tensor product of LZ and RZ; i.e., the completion in the least cross norm of the algebraic tensor product.
As a consequence of the universal properties of (ε, S GP ) and (ε, S AB ), it is trivial that (ε, S AB∩GP ) is the universal abelian group compactification of S. Some other facts about AB ∩ GP are collected in the next result.Also, see [3].
(ii) AB ∩ GP is the closed linear span of the set of all continuous characters of S.
Proof.The facts that S AB ∩ MD and S AB ∩ SD are abelian groups and that (ε, S AB ∩ GP ) is universal with respect to the property of being an abelian group imply that AB ∩ MD = AB ∩ GP = AB ∩ SD ⊆ SAP , where the latter containment is obtained from the Ellis' (joint continuity) theorem [4].Furthermore, the other condition in the definition of AB, i.e., f (st) = f (ts) is automatically deduced from f (stu) = f (sut) and the fact that f ∈ SAP .The observation that the dual mapping of ε from C(S AB ∩ GP ) onto AB ∩ GP establishes a one-to-one correspondence between the continuous characters of S AB ∩ GP and those of S and using the Peter-Weyl theorem, [5,Thm. 22.17], for C(S AB ∩ GP ) imply that AB ∩ GP is the closed linear span of the continuous characters of S.
Examples and Remarks 3.5.(i) For all right zero and left zero semigroups, it is simple to verify that AB = C (i.e., consists of the constant functions only) and that BD = C(S).Also, for all groups BD = C.
(ii) Consider the discrete semigroup S = {a, b, c, d}, with multiplication given by: a as a left identity, b and c be as left zeros, and ds = c for all s ∈ S (see [1, 1.1.7](viii) It should be mentioned that AB, SL, BD ∩D, and AB ∩ GP are also admissible, i.e., they are invariant under T µ for all µ in their duals [1,Cor. 4.2.7].But we guess that BD is not admissible in general.It would be desirable to investigate the inclusion BD ⊆ WAP .
(ix) Parallel to BD and also SL which are defined by right translates, we have the analogous spaces defined by left translates.It is a matter of fact that the left and right notations do not change the structure of SL (see Proposition 3.3).A natural question that arises is whether they do not change BD.In our opinion, there is a close tie between the latter question and the inclusion BD ⊆ WAP .See (viii).
(x) It is obvious that the SL-compactification of the direct product of two semitopological semigroups is isomorphic (in the sense of [1,Sec. 5.2]) to the direct product of their SL-compactificatons.A similar fact holds for the (AB ∩ GP )-compactification; (more generally, (AB ∩GP )-compactification, roughly speaking, passes through semidirect products.See [1,Lem. 5.2.3]).

( v )
Similar to what we have preceding to Proposition 3.3, using the Lawson's theorem,[8], one may show that for abelian semigroups BD ∩ WAP = SL = BD ∩ AP .Thus, for semilattices, SL = AP .(vi)The equality BD ∩ MD = LZ can be easily demonstrated from the fact that all left simple bands are left zero semigroups.Similarly, BD ∩ SD = RZ .Also, we trivially have BD ∩ GP = BD ∩ SAP = C.(vii) The invariant mean on the abelian semigroup S AB induces a unique invariant mean on AB, where the uniqueness is obtained from the fact that the m-admissible subalgebras of WAP cannot have more than one invariant mean (see [1, Cor.2.3.28,Exereise 4.2.7]).A similar statement holds for SL and AB ∩ GP .But BD, in general, is not even left amenable.For example, for S = {a, b, c, d} as in part (ii), let f in BD be such that f (b) = f (c), then for each left invariant mean m on BD, f (b) = m(L b f ) = m(L c f ) = f (c) and this contradicts the choice of f .