Isomorphic groupoid \(C^\ast\)-algebras associated with different Haar systems (Q2581075)
From MaRDI portal
scientific article
Language | Label | Description | Also known as |
---|---|---|---|
English | Isomorphic groupoid \(C^\ast\)-algebras associated with different Haar systems |
scientific article |
Statements
Isomorphic groupoid \(C^\ast\)-algebras associated with different Haar systems (English)
0 references
10 January 2006
0 references
The author considers a second countable Hausdorff locally compact groupoid \(G\) [cf. \textit{H. Brandt}, Math. Ann. 96, 360--366 (1926; JFM 52.0110.09); Math. Ann. 96, 353--359 (1926; JFM 52.0145.02)], with several technical conditions attached. The base-space is the space of units, denoted by \(G^{0}\). The range and source mappings \(G \to G^{0}\) are respectively \(r:x \mapsto xx^{-1}\) and \(s:x \mapsto x^{-1}x\), assumed to be open mappings, there is an inverse for elements of \(G\) and multiplication for the set of pairs \((x,y) \in G \times G\) for which \(s(x) = r(y)\). For units \(u\) and \(v\), \(G^{u} = \{ x \in G:r(x) = u\}, G_{v} = \{ x \in G : s(x) = u \}\) are fibres over the base space. This notation extends to sets in \(G^{0}\). Two units \(u\) and \(v\) are considered as equivalent if \(G^{u}_{v}= G^{u} \cap G_{v}\) is nonempty. This equivalence class, denoted \([u]\), is the set of \(v \in G^{0}\) such that there exists a \(g\) with \(r(g) = u, s(g) = v\). These equivalence classes are called orbits. The orbit space is assumed to be Hausdorff. The principal groupoid, also Hausdorff, associated with this equivalence relation is defined to be the graph \((r,s)(G)\) in \(G^{0} \times G^{0}\), (\(r,s\)) being an injective (fibre bundle) morphism. The groupoid is called proper if the inverse image of a compact subset of \(G^{0} \times G^{0}\) is compact. A continuous (left) Haar system, if it exists, is based on a family \((\lambda^{u}):u \in G^{0}\) of measures on \(G\) having as support the fibres \(G^{u}\), such that \(u \mapsto \int{f(y) \,d\lambda^{u}(y)}\) is continuous for \(f \in C_{c}(G)\) and left invariant in that left multiplication by \(g \in G\) takes \(\lambda^{s(g)}\) to \(\lambda^{r(g)}\). One defines a regular Borel measure \(\lambda\) on \(G\) by \(\int{f(y)\,d\lambda(y)} = \int {\int{f(y)\,d\lambda^{u}(y)}d\mu(u)}\) where \(f \geq 0\) is Borel and the measure \(\mu\) on \(G^{0}\) is quasi-invariant with respect to (\(G,\lambda\)), i.e., invariant under the inverse mapping on \(G\). The Haar system reduces to Haar measure when \(G\) is a group. The measure \(\lambda\) is said to be induced by \(\mu\), in the sense of quasi-invariant measures for group representations induced from a subgroup. The topology of \(G\) does not determine an essentially unique left Haar system. A representation of \(G\) is determined by a measurable Hilbert bundle of Hilbert spaces over \(G^{0}\) with quasi-invariant measure \(\mu\) and linear isometries from \(H^{s(g)} \to H^{{r(g)}}\). One obtains unitary representations of a convolution {*}-algebra \(C_{c}(G,\lambda)\) by integration [cf. \textit{J. Renault}, J. Oper. Theor. 18, 67--97 (1987; Zbl 0659.46058)], analogous to the construction of representations of \(L^{1}\)-group algebras from unitary representations of the groups. The linear mappings \(H^{s(g)} \to H^{{r(g)}}\) become unitary operators in the \(L^{2}\) space of sections of the bundle. The groupoid C*-algebras are constructed from representations of \(C_{c}(G,\lambda)\); for example the universal C*-algebra \(C^{*}(G,\lambda)\) the completion of the universal representation which is a subalgebra of \(M^{*}(G,\lambda)\) as defined by \textit{A. Ramsay} and \textit{M. Walter} [J. Funct. Anal. 148, No. 2, 314--367 (1997; Zbl 0901.43002)]. In \(\S3\) the author applies, for subsequent use, results of J. Renault [loc. cit.] on Haar systems of measures. Renault's locally compact \(G\)-space \(X\) is defined as having a projection \(p\) of \(X\) onto \(G^{0}\) and a `semi-direct' product for sets of pairs \((g,x)\) for which \(s(g) = p(x)\). For an open mapping \(\pi:X \to Y\) he considers a suitable family of measures \(\rho_{y}\) with supports \(\pi^{-1}(y)\) in \(X\). He calls \(\pi\) \(G\)-equivariant if \(g\rho_{y} = \rho_{gy}\) for all couples \((g,y)\). An equivariant \(G\)-system \(\rho\) induces a quotient system on the `orbit spaces' \(X/G\) onto \(Y/G\). To each \(u \in G^{0}\) Renault associates a left Haar measure \(\beta^{u}_{u}\) supported by the isotropy group \(G^{u}_{u}\). One can then construct unique measures \(\beta^{u}_{v}\) on the \(G^{u}_{v}\). There is also a system of quasi-invariant measures \(\mu^{[u]}\) with supports \([u]\). There is thus a unique Haar system \(\alpha^{u} = \delta_{u}\times \mu^{[u]}\) with supports \(\{ u \} \times [u]\) such that \(\lambda^{u} = \int \beta^{s}_{v}\, d\alpha^{u}(u,v)\) for all equivalent pairs \((u,v) \in G^{0}\). Seeing that the open morphism \(g \mapsto (r(g)s(g))\) from \(G\) to its principal groupoid is equivariant the author applies Renault's results and gets a `decomposition' for a Haar system over the equivalence relation (or principal groupoid) associated to \(G\). For the remainder of the article the author chooses a continuous Haar system and assumes that \(G\) has an open Hausdorff orbit space which is proper (as defined in the article). She assumes also that there is an \(F \subset G^{0}\) which crosses each orbit only once. The existence of such a set when \(r\) is an open mapping (as ensured by the quasi-invariance) follows as a generalisation of \textit{G. Mackey}'s condition [Ann. Math. 55, 101--139 (1952; Zbl 0046.11601)], for representations of a group, induced from a subgroup, on the coset spaces of the subgroup. Using \(C_{\sigma_{F}}\) instead of \(C_{c}\) she associates a C*-algebra \(C^{*}_{\sigma_{F}}(G,\nu)\) to each cross-section \(\sigma_{F}: G^{(0)} \to G^{F}\) and proves that, dependent on having a suitable Hausdorff quotient topology induced on \(G^{F}_{F}\), if \(\sigma_{F}(K)\) is relatively compact for all compact \(K\) then for two Haar systems \(\lambda_{1}\) and \(\lambda_{2}\) the C*-algebras \(M_{\sigma_{F}}{(G,\lambda_{1})}\) and \(M_{\sigma_{F}}{(G,\lambda_{2})}\) are isomorphic. Also for a \((G,\lambda)\) as above, \(C^{*}(G,\lambda)\subset M^{*}_{\sigma_{F}}(G, \lambda) \subset M^{*}(G,\lambda)\). In the transitive (i.e., a single orbit) case these C*-algebras are known to coincide and the author shows this is true also for locally transitive groupoids, i.e., where all orbits are open in \(G^{0}\). A definition is given for a Hilbert bundle determined by the Haar system (\(G,\{ \mu^{[u]}\}\)) to have a continuous basis. Considering two Haar systems with continuous decompositions over their principal groupoids, if the Hilbert bundles have continuous bases then the corresponding universal C*-algebras are isomorphic. The paper seems to be presented for people who know it all already and \(\S 3\) is particularly hard to read. The reviewer suspects that a better explanation of this section is to appear in a yet unpublished book, cited in the bibliography, by P. Muhly.
0 references
locally compact groupoid
0 references
Haar system
0 references
equivalence relation
0 references
orbit
0 references
induced representation
0 references
\(C^*\)-algebra
0 references