A not so simple local multiplier algebra (Q2499257): Difference between revisions

Let \(A\) be a C*-algebra, \(M(A)\) its multiplier algebra. All ideals in the article are assumed to be closed and two-sided. An ideal \(I \in M(A)\) is called essential if \(I \cap A = 0\) implies \(I = 0\). If \(I\) and \(J\) are essential, then the inclusion \(J \subset I\) restricts to a *-monomorphism \(M(I)\to M(J)\). The local multiplier algebra \(M_{loc}(A)\) is defined to be the inductive limit of the \(M(I)\) following these mappings; one thus has a localisation to a sequence of essential ideals and then an enlargement containing all the multipliers (cf.\ \textit{G.\,K.\thinspace Pedersen} [Invent.\ Math.\ 45, 299--305 (1978; Zbl 0376.46040)] who, while investigating extensions of derivations, defined the local algebra, denoting it by \(M^{\infty}(A)\), and developed its ideal structure). The extension of a C*-algebra \(A\) to \(M(A)\) is mainly functional-analytic in nature, whereas the extension to \(M_{loc}(A)\) is mainly algebraic. The extensions coincide when \(A\) is simple and prime. When \(A\) is Abelian, \(M(C_{0}(X)) = C_{b}(X)\), so here the multipliers of closed ideals are ``local'' in that they are algebras of bounded continuous functions on open bounded subsets of \(X\). The aim of the present article, in response to a question by Pedersen [loc.\,cit.], is to construct unital primitive AF-algebras \(A\) such that \(M_{loc}(M_{loc}(A)) \neq M_{loc}(A)\). Instead of the usual K-theoretic characterisation of AF-algebras, the authors use the Abelian monoid \(V(A)\) of Murray--von Neumann equivalence classes of projections in \(M_{loc}(A)\). This gives better control of the order properties of the ideals. The monoid has a natural order structure; hereditary sub-monoids are called order-ideals, upward directed hereditary subsets are called intervals. A complemented interval in a fixed interval \(D\) is taken to be complemented in \(kD\) for some integer \(k\geq 1\). The set of complemented intervals is denoted by \(\Lambda(V,D)\) and it is a monoid for a naturally defined sum of intervals. The monoid of local intervals \(\Lambda_{loc}(V,1)\) is the downward inductive limit of sequences \(\Lambda(V,D_{N})\), where the \(N\) form a directed set of essential ideals. \textit{G.\,A.\thinspace Elliott} [C.\,R.\thinspace Math.\ Rep.\ Acad.\ Sci.\ Canada 9, 225--230 (1987; Zbl 0637.46062)] used \(K_{0}(M(A))\) to determine an order structure for ideals of AF-algebras. The authors in the present paper use monoid-based results of \textit{S.\,Zhang} [J.~Oper.\ Theory 24, No.\,2, 209--226 (1990; Zbl 0747.46043)], \textit{K.\,R.\thinspace Good\-earl} [K-Theory 10, No.\,5, 419--489 (1996; Zbl 0857.46044)] and \textit{F.\,Perera} [Can.\ J.\ Math.\ 53, No.\,3, 592--630 (2001; Zbl 0976.46040)]. One can describe the ideal structure of \(M(A)\) in terms of the cancellative monoid \(V(M(A))\). The authors rely on the construction from their article [Math.\ Proc.\ Camb.\ Philos.\ Soc.\ 126, No.\,3, 555--564 (1999; Zbl 0945.46050)] which demonstrated the existence of a countable subgroup of the lattice structure of the space of real continuous functions, with an order relation, on an infinite compact metrizable set. The lattice of ideals of \(M(A)\) is isomorphic to the lattice of order ideals in \(V(M(A))\). The authors [loc.\,cit.] extended this result by using their localisation procedure to show that \(V(M_{loc}(A))\) is isomorphic to the monoid of local intervals. According to the authors [loc.\,cit.], there is a countable cancellative monoid \(M\) which, with an algebraic order, leads to a localised monoid \(M'\), not necessarily cancellative, representing \(V(M_{loc}(A))\). There is a unique minimal order-ideal \(J\) in \(M'\) such that \(M_{loc}(M_{loc}(J)) = M_{loc}(J)\). However, \(J\) does not have a countable approximate identity and the projections in \(J\) fail to satisfy cancellation, so the structure of the projections in \(M(J)\) is generally undetermined. Neveretheless, they construct a sequence of projections in \(M_{loc}(A)\) converging strictly in the \(J\)-topology to a projection in \(M(J)\backslash M_{loc}(A)\). Hence \(M_{loc}(M_{loc}(A)) \neq M_{loc}(A)\). The proofs are very technical. Yet, in the last section, following Pedersen's [loc.\,cit.]\ more natural approach, \(M_{loc}(A)\) is expressed as the inductive limit of fundamental sequences of ideals. The authors assert that this could lead to further examples of AF algebras for which \(M_{loc}(M_{loc}(A)) \neq M_{loc}(A)\).