Isomorphic groupoid C *   -algebras associated with different Haar systems

Mădălina Roxana BuneciThis work was partly supported by the MEC-CNCSIS grant At127/2004 and by the Postdoctoral Training Program HPRN-CT-2002-0277. University Constantin Brâncuşi, Târgu-Jiu

Abstract
We shall consider a locally compact groupoid endowed with a Haar system ν   and having proper orbit space. We shall associated to each appropriated cross section σ : G ( 0 ) → G F   for d F : G F → G ( 0 )   (where F   is a Borel subset of G ( 0 )   meeting each orbit exactly once) a C *   -algebra M σ * ( G , ν )   . We shall prove that the C *   -algebras M σ * ( G , ν i )   i = 1 , 2   associated with different Haar systems are *   -isomorphic AMS 2000 Subject Classification: 22A22, 43A22, 43A65, 46L99.
Key Words: locally compact groupoid, C *   -algebra, *   -isomorphism.

1 Introduction

The C *   -algebra of a locally compact groupoid was introduced by J. Renault in [8] . The construction extends the case of a group: the space of continuous functions with compact support on groupoid is made into a *   -algebra and endowed with the smallest C *   -norm making its representations continuous. In order to define the convolution on groupoid one needs to assume the existence of a Haar system which is an analogue of Haar measure on a group. Unlike the case for groups, Haar systems need not to be unique. A result of Paul Muhly, Jean Renault and Dana Williams establishes that the C *   -algebras of G   associated with two Haar systems are strongly Morita equivalent (Theorem 2.8/p. 10 [3] ). If the groupoid is transitive they have proved that if G   is transitive then the C *   -algebra of G   is isomorphic to C * ( H ) ⊗ K ( L 2 ( μ ) )   , where H   is the isotropy group G u u   at any unit u ∈ G ( 0 )   , μ   is an essentially unique measure on G ( 0 )   , C * ( H )   denotes the group C *   -algebra of H   , and K ( L 2 ( μ ) )   denotes the compact operators on L 2 ( μ )   (Theorem 3.1/p. 16 [3] ). Therefore the C *   -algebras of a transitive groupoid G   associated with two Haar systems are *   -isomorphic.
In [7] Arlan Ramsay and Martin E. Walter have associated to a locally compact groupoid G   a C *   algebra denoted M * ( G , ν )   . They have considered the universal representation ω   of C * ( G , ν )   -the usual C *   -algebra associated to a Haar system ν = { ν u , u ∈ G ( 0 ) }   (constructed as in [8] ) . Since every cyclic representation of C * ( G , ν )   is the integrated form of a representation of G   , it follows that ω   can be also regarded as a representation of ℬ c ( G )   , the space of compactly supported Borel bounded function on G   . Arlan Ramsay and Martin E. Walter have used the notation M * ( G , ν )   for the operator norm closure of ω ( ℬ c ( G ) )   . Since ω   is an *   -isomorphism on C * ( G , ν )   , we can regarded C * ( G , ν )   as a subalgebra of M * ( G , ν )   .
We shall assume that the orbit space of the groupoid G   is proper and we shall choose a Borel subset F   of G ( 0 )   meeting each orbit exactly once and such that F ∩ [ K ]   has a compact closure for each compact subset K   of G ( 0 )   . For each appropriated cross section σ : G ( 0 ) → G F   for d F : G F → G ( 0 )   , d F ( x ) = d ( x )   , we shall construct a C *   -algebra M σ * ( G , ν )   which can be viewed as a subalgebra of M * ( G , ν )   . If ν 1 = { ν 1 u , u ∈ G ( 0 ) }   and ν 2 = { ν 2 u , u ∈ G ( 0 ) }   are two Haar systems on G   , we shall prove that the C *   -algebras M σ * ( G , ν 1 )   and M σ * ( G , ν 2 )   are *   -isomorphic.
For a transitive (or more generally, a locally transitive) groupoid G   we shall prove that the C *   -algebras C * ( G , ν )   , M * ( G , ν )   and M σ * ( G , ν )   coincide.
For a principal proper groupoid G   , we shall prove that C * ( G , ν ) ⊂ M σ * ( G , ν ) ⊂ M * ( G , ν ) .   Let π : G ( 0 ) → G ( 0 ) / G   be the quotient map and let ν i = { ɛ u × μ i π ( u ) , u ∈ G ( 0 ) }   , i = 1 , 2   be two Haar systems on the principal proper groupoid G   . We shall also prove that if the Hilbert bundles determined by the systems of measures { μ i u ˙ } u ˙   have continuous bases in the sense of Definition  24 , then *   -isomorphism between M σ * ( G , ν 1 )   and M σ * ( G , ν 2 )   can be restricted to a *   -isomorphism between C * ( G , ν 1 )   and C * ( G , ν 2 )   .
For establishing notation, we include some definitions that can be found in several places (e.g. [8] , [4] ). A groupoid is a set G   endowed with a product map ( x , y ) → x y [ : G ( 2 ) → G ]   where G ( 2 )   is a subset of G × G   called the set of composable pairs, and an inverse map x → x − 1 [ : G → G ]   such that the following conditions hold:
( 1 )   If ( x , y ) ∈ G ( 2 )   and ( y , z ) ∈ G ( 2 )   , then ( x y , z ) ∈ G ( 2 )   , ( x , y z ) ∈ G ( 2 )   and ( x y ) z = x ( y z )   .
( 2 )   ( x − 1 ) − 1 = x   for all x ∈ G   .
( 3 )   For all x ∈ G   , ( x , x − 1 ) ∈ G ( 2 )   , and if ( z , x ) ∈ G ( 2 )   , then ( z x ) x − 1 = z   .
( 4 )   For all x ∈ G   , ( x − 1 , x ) ∈ G ( 2 )   , and if ( x , y ) ∈ G ( 2 )   , then x − 1 ( x y ) = y   .
The maps r   and d   on G   , defined by the formulae r ( x ) = x x − 1   and d ( x ) = x − 1 x   , are called the range and the source maps. It follows easily from the definition that they have a common image called the unit space of G   , which is denoted G ( 0 )   . Its elements are units in the sense that x d ( x ) = r ( x ) x = x   .
Units will usually be denoted by letters as u , v , w   while arbitrary elements will be denoted by x , y , z   . It is useful to note that a pair ( x , y )   lies in G ( 2 )   precisely when d ( x ) = r ( y )   , and that the cancellation laws hold (e.g. x y = x z   iff y = z   ) .   The fibres of the range and the source maps are denoted G u = r − 1 ( { u } )   and G v = d − 1 ( { v } )   , respectively. More generally, given the subsets A   , B   ⊂ G ( 0 )   , we define G A = r − 1 ( A )   , G B = d − 1 ( B )   and G B A = r − 1 ( A ) ∩ d − 1 ( B )   . The reduction of G   to A ⊂ G ( 0 )   is G | A = G A A   . The relation u ~ v   iff G v u ≠ φ   is an equivalence relation on G ( 0 )   . Its equivalence classes are called orbits and the orbit of a unit u   is denoted [ u ]   . A groupoid is called transitive iff it has a single orbit. The quotient space for this equivalence relation is called the orbit space of G   and denoted G ( 0 ) / G   . We denote by π : G ( 0 ) → G ( 0 ) / G , π ( u ) = u ˙   the quotient map. A subset of G ( 0 )   is said saturated if it contains the orbits of its elements. For any subset A   of G ( 0 )   , we denote by [ A ]   the union of the orbits [ u ]   for all u ∈ A   .
A topological groupoid consists of a groupoid G   and a topology compatible with the groupoid structure. This means that:
( 1 ) x → x − 1 [ : G → G ]   is continuous.
( 2 )   ( x , y ) [ : G ( 2 ) → G ]   is continuous where G ( 2 )   has the induced topology from G × G   .
We are exclusively concerned with topological groupoids which are second countable, locally compact Hausdorff. It was shown in [6] that measured groupoids may be assume to have locally compact topologies, with no loss in generality.
If X   is a locally compact space, C c ( X )   denotes the space of complex-valuated continuous functions with compact support. The Borel sets of a topological space are taken to be the σ   -algebra generated by the open sets. The space of compactly supported bounded Borel function on X   is denoted by ℬ c ( X )   . For a locally compact groupoid G   , we denote by G ′ = ( x ∈ G : r ( x ) = d ( x ) )   the isotropy group bundle of G   . It is closed in G   .
Let G   be a locally compact second countable groupoid equipped with a Haar system, i.e. a family of positive Radon measures on G   , { ν u , u ∈ G ( 0 ) }   , such that 1) For all u ∈ G ( 0 )   , s u p p ( ν u ) = G u   .
2) For all f ∈ C c ( G )   , u → ∫ f ( x ) d ν u ( x ) [ : G ( 0 ) → C ]   is continuous.
3) For all f ∈ C c ( G )   and all x ∈ G   , ∫ f ( y ) d ν r ( x ) ( y ) = ∫ f ( x y ) d ν d ( x ) ( y )   As a consequence of the existence of continuous Haar systems, r , d : G → G ( 0 )   are open maps ([10] ). Therefore, in this paper we shall always assume that r : G → G ( 0 )   is an open map If μ   is a measure on G ( 0 )   , then the measure ν = ∫ ν u d μ ( u )   , defined by ∫ f ( y ) d ν ( y ) = ∫ ( ∫ f ( y ) d ν u ( y ) ) d μ ( u ) , f ≥ 0 Borel   is called the measure on G   induced by μ   . The image of ν   by the inverse map x → x − 1   is denoted ν − 1   . μ   is said quasi-invariant if its induced measure ν   is equivalent to its inverse ν − 1   A measure belongings to the class of a quasi-invariant measure is also quasi-invariant. We say that the class is invariant.
If μ   is a quasi-invariant measure on G ( 0 )   and ν   is the measure induced on G   , then the Radon-Nikodym derivative Δ = d ν d ν − 1   is called the modular function of μ   .
In order to define the C *   -algebra of a groupoid the space of continuous functions with compact support on groupoid is made into a *   -algebra and endowed with the smallest C *   -norm making its representations continuous. For f   , g ∈ C c ( G )   the convolution is defined by: f * g ( x ) = ∫ f ( x y ) g ( y − 1 ) d ν d ( x ) ( y )   and the involution by f * ( x ) = f ( x − 1 ) ¯ .   Under these operations, C c ( G )   becomes a topological *   -algebra.
A representation of C c ( G )   is a *   -homomorphism from C c ( G )   into ℬ ( H )   , for some Hilbert space H   , that is continuous with respect to the inductive limit topology on C c ( G )   and the weak operator topology on ℬ ( H )   . The full C *   -algebra C * ( G )   is defined as the completion of the involutive algebra C c ( G )   with respect to the full C *   -norm ∥ f ∥ = sup ∥ L ( f ) ∥   where L   runs over all non-degenerate representation of C c ( G )   which are continuous for the inductive limit topology.
Every representation ( μ , G ( 0 ) * ℋ , L )   (see Definition 3.20/p.68 [4] ) of G   can be integrated into a representation, still denoted by L   , of C c ( G )   . The relation between the two representation is: 〈 L ( f ) ξ 1 , ξ 2 〉 = ∫ f ( x ) 〈 L ( x ) ξ 1 ( d ( x ) ) , ξ 2 ( r ( x ) ) 〉 Δ − 1 2 ( x ) d ν u ( x ) d μ ( u )   where f ∈   C c ( G )   , ξ 1 , ξ 2 ∈ ∫ G ( 0 ) ⊕ ℋ ( u ) d μ ( u )   .
Conversely, every non-degenerate *   -representation of C c ( G )   is obtained in this fashion (see [8] or [4] ).

2 The decomposition of a Haar system over the principal groupoid

First we present some results on the structure of the Haar systems, as developed by J. Renault in Section 1 of [9] and also by A. Ramsay and M.E. Walter in Section 2 of [7] .
In Section 1 of [9] Jean Renault constructs a Borel Haar system for G ′   . One way to do this is to choose a function F 0   continuous with conditionally support which is nonnegative and equal to 1   at each u ∈ G ( 0 ) .   Then for each u ∈ G ( 0 )   choose a left Haar measure β u u   on G u u   so the integral of F 0   with respect to β u u   is 1 .   Renault defines β v u = x β v v   if x ∈ G v u   (where x β v v ( f ) = ∫ f ( x y ) d β v v ( y )   as usual). If z   is another element in G v u   , then x − 1 z ∈ G v v   , and since β v v   is a left Haar measure on G v v   , it follows that β v u   is independent of the choice of x   . If K   is a compact subset of G   , then sup u , v β v u ( K ) < ∞   . Renault also defines a 1-cocycle δ   on G   such that for every u ∈ G ( 0 )   , δ | G u u   is the modular function for β u u   . δ   and δ − 1 = 1 / δ   are bounded on compact sets in G   .
Let R = ( r , d ) ( G ) = { ( r ( x ) , d ( x ) ) , x ∈ G }   be the graph of the equivalence relation induced on G ( 0 )   . This R   is the image of G   under the homomorphism ( r , d )   , so it is a σ   -compact groupoid. With this apparatus in place, Renault describes a decomposition of the Haar system { ν u , u ∈ G ( 0 ) }   for G   over the equivalence relation R   (the principal groupoid associated to G   ). He proves that there is a unique Borel Haar system α   for R   with the property that ν u = ∫ β t s d α u ( s , t ) for all u ∈ G ( 0 ) .   In Section 2 [7] A. Ramsay and M.E. Walter prove that sup u α u ( ( r , d ) ( K ) ) < ∞ , for all compact K ⊂ G   For each u ∈ G ( 0 )   the measure α u   is concentrated on { u } × [ u ]   . Therefore there is a measure μ u   concentrated on [ u ]   such that α u = ɛ u × μ u   , where ɛ u   is the unit point mass at u   . Since { α u , u ∈ G ( 0 ) }   is a Haar system, we have μ u = μ v   for all ( u , v ) ∈ R   , and the function u → ∫ f ( s ) μ u ( s )   is Borel for all f ≥ 0   Borel on G ( 0 )   . For each u   the measure μ u   is quasi-invariant (Section 2 [7] ). Therefore μ u   is equivalent to d * ( v u )   (Lemma 4.5/p. 277 [5] ).
If η   is a quasi-invariant measure for { ν u , u ∈ G ( 0 ) }   , then η   is a quasi-invariant measure for { α u , u ∈ G ( 0 ) }   . Also if Δ R   is the modular function associated to { α u , u ∈ G ( 0 ) }   and η   , then Δ = δ Δ R ∘ ( r , d )   can serve as the modular function associated to { ν u , u ∈ G ( 0 ) }   and η   .
Since μ u = μ v   for all ( u , v ) ∈ R   , the system of measures { μ u } u   may be indexed on the elements of the orbit space G ( 0 ) / G   .
Definition 1 We shall call the pair of the system of measures ( { β v u } ( u , v ) ∈ R , { μ u ˙ } u ˙ ∈ G ( 0 ) / G )   (described above) the decomposition of the Haar system { ν u , u ∈ G ( 0 ) }   over the principal groupoid associated to G   . Also we shall call δ   the 1   -cocycle associated to the decomposition.
Remark 2 Let us note that the system of measures { β v u }   and the 1   -cocycle in the preceding definition do not depend on the Haar system.
Lemma 3 Let G   be a locally compact second countable groupoid with the bundle map r | G ′   of G ′   open. Let { ν u , u ∈ G ( 0 ) }   be a Haar system on G   and ( { β v u } , { μ u ˙ } )   its decomposition over the principal groupoid associated to G   .
Then for each f ∈ C c ( G )   the function x → ∫ f ( y ) d β d ( x ) r ( x ) ( y )   is continuous on G   .
Proof. By Lemma 1.3/p. 6 [9] , for each f ∈ C c ( G )   the function u → ∫ f ( y ) d β u u ( y )   is continuous .   Let x ∈ G   and ( x i ) i   be a sequence in G   converging to x   . Let f ∈ C c ( G )   and let g   be a continuous extension on G   of y → f ( x y ) [ : G d ( x ) → C ]   . Let K   be the compact set ( { x , x i , i = 1 , 2 , . . } − 1 s u p p ( f ) ∪ s u p p ( g ) ) ∩ r − 1 ( { d ( x ) , d ( x i ) , i = 1 , 2 , . . . } ) .   We have | ∫ f ( y ) d β d ( x ) r ( x ) ( y ) − ∫ f ( y ) d β d ( x i ) r ( x i ) ( y ) |   = | ∫ f ( x y ) d β d ( x ) d ( x ) ( y ) − ∫ f ( x i y ) d β d ( x i ) d ( x i ) ( y ) |   = | ∫ g ( y ) d β d ( x ) d ( x ) ( y ) − ∫ f ( x i y ) d β d ( x i ) d ( x i ) ( y ) |   ≤ | ∫ g ( y ) d β d ( x ) d ( x ) ( y ) − ∫ g ( y ) d β d ( x i ) d ( x i ) ( y ) | +   + | ∫ g ( y ) d β d ( x i ) d ( x i ) ( y ) − ∫ f ( x i y ) d β d ( x i ) d ( x i ) ( y ) |   ≤ | ∫ g ( y ) d β d ( x ) d ( x ) ( y ) − ∫ g ( y ) d β d ( x i ) d ( x i ) ( y ) | +   + sup y ∈ G d ( x i ) d ( x i ) | g ( y ) − f ( x i y ) | β d ( x i ) d ( x i ) ( K )   A compactness argument shows that sup y ∈ G d ( x i ) d ( x i ) | g ( y ) − f ( x i y ) |   converges to 0   . Also | ∫ g ( y ) d β d ( x ) d ( x ) ( y ) − ∫ g ( y ) d β d ( x i ) d ( x i ) ( y ) |   converges to 0   because the function u → ∫ f ( y ) d β u u ( y )   is continuous .   Hence | ∫ f ( y ) d β d ( x ) r ( x ) ( y ) − ∫ f ( y ) d β d ( x i ) r ( x i ) ( y ) |   converges to 0 .  
Definition 4 A locally compact groupoid G   is proper if the map ( r , d ) : G → G ( 0 ) × G ( 0 )   is proper (i.e. the inverse map of each compact subset of G ( 0 ) × G ( 0 )   is compact).(Definition 2.1.9/p. 37 [1] ).
Throughout this paper we shall assume that G   is a second countable locally compact groupoid for which the orbit space is Hausdorff and the map ( r , d ) : G → R , ( r , d ) ( x ) = ( r ( x ) , d ( x ) )   is open, where R   is endowed with the product topology induced from G ( 0 ) × G ( 0 )   . Therefore R   will be a locally compact groupoid. The fact that R   is a closed subset of G ( 0 ) × G ( 0 )   and that it is endowed with the product topology is equivalent to the fact R   is a proper groupoid.
Throughout this paper by a groupoid with proper orbit space we shall mean a groupoid G   for which the orbit space is Hausdorff and the map ( r , d ) : G → R , ( r , d ) ( x ) = ( r ( x ) , d ( x ) )   is open, where R   is endowed with the product topology induced from G ( 0 ) × G ( 0 )   .
Proposition 5 Let G   be a second countable locally compact groupoid with proper orbit space. Let { ν u , u ∈ G ( 0 ) }   be a Haar system on G   and ( { β v u } , { μ u ˙ } )   its decomposition over the principal groupoid associated to G   . Then .for each g ∈ C c ( G ( 0 ) )   , the map u → ∫ g ( v ) d μ π ( u ) ( v )   is continuous.
Proof. Let g ∈ C c ( G ( 0 ) )   and u 0 ∈ G ( 0 )   . Let K 1   be a compact neighborhood of u 0   and K 2   be the support of g   . Since G   is locally compact and ( r , d )   is open from G   to ( r , d ) ( G )   , there is a compact subset K   of G   such that ( r , d ) ( K )   contains ( K 1 × K 2 ) ∩ ( r , d ) ( G )   . Let F 1 ∈ C c ( G )   be a nonnegative function equal to 1   on a compact neighborhood U   of K   . Let F 2   ∈ C c ( G )   be a function which extends to G   the function x → F 1 ( x ) / ∫ F 1 ( y ) d β d ( x ) r ( x ) ( y )   , x ∈ U   . We have ∫ F 2 ( y ) d β v u ( y ) = 1   for all ( u , v ) ∈ ( r , d ) ( K )   . Since for all u ∈ K 1   ,
∫ g ( v ) d μ π ( u ) ( v ) = ∫ g ( v ) ∫ F 2 ( y ) d β v u ( y ) d μ π ( u ) ( v )
= ∫ g ( d ( y ) ) F 2 ( y ) d ν u ( y ) ,
it follows that u → ∫ g ( v ) d μ π ( u ) ( v )   is continuous at u 0 .  
Remark 6 Let G   be a locally compact second countable groupoid with proper orbit space. Let { ν u , u ∈ G ( 0 ) }   be a Haar system on G   and ( { β v u } , { μ u ˙ } )   be its decomposition over the associated principal groupoid. If μ   is a quasi-invariant probability measure for the Haar system, then μ 1 = ∫ μ π ( u ) d μ ( u )   is a Radon measure which is equivalent to μ .   Indeed, let f ≥ 0   Borel on G ( 0 )   such that μ ( f ) = 0   . Since μ   is quasi-invariant, it follows that for μ a . a . u   ν u ( f ∘ d ) = 0   , and since μ π ( u )   is equivalent to d * ( v u )   , it results μ π ( u ) ( f ) = 0   for μ   a . a . u   .
Conversely, if μ 1 ( f ) = 0   , then μ π ( u ) ( f ) = 0   for μ   a . a . u   , and therefore ν u ( f ∘ d ) = 0   . Thus the quasi-invariance of μ   implies μ ( f ) = 0   . Thus each Radon quasi-invariant measure is equivalent to a Radon measure of the form ∫ μ u ˙ d μ ~ ( u ˙ )   , where μ ~   is a probability measure on the orbit space G / G ( 0 ) .  

3 A C *   -algebra associated to a locally compact groupoid with proper orbit space

Let G   be a locally compact second countable groupoid with proper orbit space. Let π : G ( 0 ) → G ( 0 ) / G   be the quotient map. Since the quotient space is proper, G ( 0 ) / G   is Hausdorff. Let us assume that the range map r   is open. As a consequence, the map π   is open. Applying Lemma 1.1 [2] to the locally compact second countable spaces G ( 0 )   and G ( 0 ) / G   and to the continuous open surjection π : G ( 0 ) → G ( 0 ) / G   , it follows that there is a Borel set F   in G ( 0 )   such that:
  • 1. F   contains exactly one element in each orbit [ u ] = π − 1 ( π ( u ) )   .
  • 2. For each compact subset K   of G ( 0 )   , F ∩ [ K ] = F ∩ π − 1 ( π ( K ) )   has a compact closure.
For each unit u   let us define e ( u ) = F ∩ [ u ]   ( e ( u )   is the unique element in the orbit of u   contained in F   ). For each Borel subset B   of G ( 0 )   , π   is continuous and one-to-one on B ∩ F   and hence π ( B ∩ F )   is Borel in G ( 0 ) / G   . Therefore the map e : G ( 0 ) → G ( 0 )   is Borel (for each Borel subset B   of G ( 0 )   , e − 1 ( B ) = [ B ∩ F ] = π − 1 ( π ( B ∩ F ) )   is Borel in G ( 0 )   ). Also for each compact subset K   of G ( 0 )   , e ( K )   has a compact closure because e ( K ) ⊂ F ∩ [ K ]   .
Since the orbit space G ( 0 ) / G   is proper the map ( r , d ) : G → R , ( r , d ) ( x ) = ( r ( x ) , d ( x ) )   is open and R   is closed in G ( 0 ) × G ( 0 )   . Applying Lemma 1.1 [2] to the locally compact second countable spaces G   and R   and to the continuous open surjection ( r , d ) : G → R   , it follows that there is a regular cross section σ 0 : R → G   . This means that σ 0   is Borel, ( r , d ) ( σ 0 ( u , v ) ) = ( u , v )   for all ( u , v ) ∈ R   , and σ 0 ( K )   is relatively compact in G   for each compact subset K   of R   .
Let us define σ : G ( 0 ) → G F   by σ ( u ) = σ 0 ( e ( u ) , u )   for all u   . It is easy to note that σ   is a cross section for d : G F → G ( 0 )   and σ ( K )   is relatively compact in G   for all compact K ⊂ G ( 0 )   .
Replacing σ   by v → σ ( e ( v ) ) − 1 σ ( v )   we may assume that σ ( e ( v ) ) = e ( v )   for all v   . Let us define q : G → G F F   by q ( x ) = σ ( r ( x ) ) x σ ( d ( x ) ) − 1 , x ∈ G .   Let ν = { ν u : u ∈ G ( 0 ) }   be a Haar system on G   and let ( { β v u } , { μ u ˙ } )   be its decompositions over the principal groupoid . Let δ   be the 1   -cocycle associated to the decomposition.
Let us denote by ℬ σ ( G )   the linear span of the functions of the form x → g 1 ( r ( x ) ) g ( q ( x ) ) g 2 ( d ( x ) )   where g 1 , g 2   are compactly supported Borel bounded function on G ( 0 )   and g   is a Borel bounded function on G F F   such that if S   is the support of g   , then the closure of σ ( K 1 ) − 1 S σ ( K 2 )   is compact in G   for all compact subsets K 1   , K 2   of G ( 0 )   . ℬ σ ( G )   is a subspace of ℬ c ( G )   , the space of compactly supported Borel bounded function on G   .
If f 1   , f 2 ∈ ℬ σ ( G )   are defined by
f 1 ( x ) = g 1 ( r ( x ) ) g ( q ( x ) ) g 2 ( d ( x ) )
f 2 ( x ) = h 1 ( r ( x ) ) h ( q ( x ) ) h 2 ( d ( x ) )
then f 1 * f 2 ( x ) = g * h ( q ( x ) ) g 1 ( r ( x ) ) h 2 ( d ( x ) ) 〈 g 2 , h 1 ¯ 〉 π ( r ( x ) )   f 1 * ( x ) = g 2 ( r ( x ) ) ¯ g ( q ( x ) − 1 ) ¯ g 1 ( d ( x ) ) ¯   Thus ℬ σ ( G )   is closed under convolution and involution.
Let ω   be the universal representation of C * ( G , ν )   the usual C *   -algebra associated to a Haar system ν = { ν u , u ∈ G ( 0 ) }   (constructed as in [8] ) . Since every cyclic representation of C * ( G , ν )   is the integrated form of a representation of G   , it follows that ω   can be also regarded as a representation of ℬ c ( G )   , the space of compactly supported Borel bounded function on G   . Arlan Ramsay and Martin E. Walter have used the notation M * ( G , ν )   for the operator norm closure of ω ( ℬ c ( G ) )   . Since ω   is an *   -isomorphism on C * ( G , ν )   , we can regarded C * ( G , ν )   as a subalgebra of M * ( G , ν )   .
Definition 7 We denote by M σ * ( G , ν )   the operator norm closure of ω ( ℬ σ ( G ) )   .
Lemma 8 Let { μ 1 u ˙ } u ˙   and { μ 2 u ˙ } u ˙   be two systems of measures on G ( 0 )   satisfying:
  • 1. s u p p ( μ i u ˙ ) = [ u ]   for all u ˙   , i = 1 , 2  
  • 2. For all compactly supported Borel bounded function f   on G ( 0 )   the function u → ∫ f ( v ) μ i π ( u ) ( v )   is bounded and Borel.
Then there is a family { U u ˙ } u ˙   of unitary operators with the following properties:
  • 1. U u ˙ : L 2 ( μ 1 u ˙ ) → L 2 ( μ 2 u ˙ )   is a unitary operator for each u ˙ ∈ G ( 0 ) / G   .
  • 2. For all Borel bounded function f   on G ( 0 )   , u → U π ( u ) ( f )   is a bounded Borel function with compact support.
  • 3. For all Borel bounded function f   on G ( 0 )   , U π ( u ) ( f ¯ ) = U π ( u ) ( f ) ¯  
Proof. Using the same argument as in [6] (p. 323) we can construct a sequence f 1 , f 2 , . . .   of real valued Borel bounded function on G ( 0 )   such that dim ( L 2 ( μ 1 u ˙ ) ) = ∞   if and only if ∥ f n ∥ 2   for n = 1 , 2 , . . .   and then { f 1 , f 2 , . . . }   gives an ortonormate basis of L 2 ( μ 1 u ˙ )   , while dim ( L 2 ( μ 1 u ˙ ) ) = k < ∞   if and only if ∥ f n ∥ 2 = 1   for n ≤ k   , and ∥ f n ∥ 2 = 0   for n > k   and then { f 1 , f 2 , . . . f k }   gives an ortonormate basis of L 2 ( μ 1 u ˙ )   . Let g 1 , g 2 , . . .   be a sequence with the same properties as f 1 . f 2 , . . .   corresponding to { μ 2 u ˙ } u ˙   . Let us define U u ˙ : L 2 ( μ 1 u ˙ ) → L 2 ( μ 2 u ˙ )   by U u ˙ ( f n ) = g n for all n   Then the family { U u ˙ } u ˙   has the required properties.
Theorem 9 Let G   be a locally compact second countable groupoid with proper orbit space. Let { ν i u , u ∈ G ( 0 ) }   , i = 1 , 2   be two Haar systems on G   . Let F   be a Borel subset of G ( 0 )   containing only one element e ( u )   in each orbit [ u ]   . Let σ : G ( 0 ) → G F   be a cross section for d : G F → G ( 0 )   with σ ( e ( v ) ) = e ( v )   for all v ∈ G ( 0 )   and σ ( K )   relatively compact in G   for all compact sets K ⊂ G ( 0 )   .
Then the C *   -algebras M σ * ( G , ν 1 )   and M σ * ( G , ν 2 )   are *   -isomorphic.
Proof. Let ( { β v u } , { μ i u ˙ } )   be the decompositions of the Haar systems over the principal groupoid . Let δ   be the 1   -cocycle associated to the decompositions, i = 1 , 2   .
We shall denote by 〈 ⋅ , ⋅ 〉 i , u ˙   the inner product of ( L 2 ( G ( 0 ) , δ ( σ ( ⋅ ) ) μ i u ˙ ) )   , i = 1 , 2   .
Let us define q : G → G F F   by q ( x ) = σ ( r ( x ) ) x σ ( d ( x ) ) − 1 , x ∈ G .   We shall define a *   -homomorphism Φ   from ℬ σ ( G )   to ℬ σ ( G )   . It suffices to define Φ   on the set of function on G   of the form x → g 1 ( r ( x ) ) g ( q ( x ) ) g 2 ( d ( x ) )   Let { U u ˙ } u ˙   be the family of unitary operators with the properties stated in Lemma  8 , associated to the systems of measures { δ ( σ ( ⋅ ) ) μ i u ˙ } u ˙   , i = 1 , 2   .
Let us define Φ   by Φ ( f ) = ( x → U π ( r ( x ) ) ( g 1 ) ( r ( x ) ) g ( q ( x ) ) U π ( d ( x ) ) ( g 2 ) ( d ( x ) ) )   where f   is defined by f ( x ) = g 1 ( r ( x ) ) g ( q ( x ) ) g 2 ( d ( x ) )   If f 1   and f 2   are defined by
f 1 ( x ) = g 1 ( r ( x ) ) g ( q ( x ) ) g 2 ( d ( x ) )
f 2 ( x ) = h 1 ( r ( x ) ) h ( q ( x ) ) h 2 ( d ( x ) )
then f 1 * f 2 ( x ) = g * h ( q ( x ) ) g 1 ( r ( x ) ) h 2 ( d ( x ) ) 〈 g 2 , h 1 ¯ 〉 1 , π ( r ( x ) )   and consequently
Φ ( f 1 * f 2 )
= g * h ( q ( x ) ) U π ( r ( x ) ) ( g 1 ) ( r ( x ) ) U π ( r ( x ) ) ( h 2 ) ( d ( x ) ) 〈 g 2 , h 1 ¯ 〉 1 , π ( r ( x ) )
= Φ ( f 1 ) * Φ ( f 2 ) .
Let η ~   be a probability measure on G ( 0 ) / G   and η i = ∫ μ i u ˙ d η ~ ( u ˙ ) , i = 1 , 2   . Let L 1   be the integrated form of a representation ( L , ℋ * G ( 0 ) , η 1 )   and L 2   be the integrated form of ( L , ℋ * G ( 0 ) , η 2 )   . Let B   be the Borel function defined by: B ( u ) = L ( σ ( u ) )   and W : ∫ G ( 0 ) ⊕ ℋ ( u ) d η 1 ( u ) → ∫ G ( 0 ) ⊕   ℋ ( e ( u ) ) d η 1 ( u )   be defined by W ( ζ ) = ( u → B ( u ) ( ζ ( u ) ) )   Since every element of L 2 ( G ( 0 ) , δ ( σ ( ⋅ ) ) μ 1 w ˙ , ℋ ( e ( w ) ) )   is a limit of linear combinations of elements u → a ( u ) ξ   with a ∈ L 2 ( G ( 0 ) , δ ( σ ( ⋅ ) ) μ 1 w ˙ )   and ξ ∈ ℋ ( e ( w ) )   , we can define a unitary operator V w ˙ : L 2 ( G ( 0 ) , δ ( σ ( ⋅ ) ) μ 1 w ˙ , ℋ ( e ( w ) ) ) → L 2 ( G ( 0 ) , δ ( σ ( ⋅ ) ) μ 2 w ˙ , ℋ ( e ( w ) ) )   by V w ˙ ( u → a ( u ) ξ ) = U w ˙ ( a ) ξ   Let V : ∫ G ( 0 ) ⊕ ℋ ( e ( u ) ) d η 1 ( u ) → ∫ G ( 0 ) ⊕   ℋ ( e ( u ) ) d η 2 ( u )   be defined by V ( ζ ) = ( u → V u ˙ ( ζ ( u ) ) )   If ζ 1 , ζ 2 ∈ ∫ G ( 0 ) ⊕   ℋ ( e ( u ) ) d η 1 ( u )   and f   is of the form f ( x ) = g 1 ( r ( x ) ) g ( q ( x ) ) g 2 ( d ( x ) ) ,   we have
〈 W L 1 ( f ) W * ζ 1 , ζ 2 〉
= ∫ ∫ g ( x ) δ ( x ) − 1 2 〈 L ( x ) A 1 ( w ˙ ) , B 1 ( w ˙ ) 〉 d β e ( w ) e ( w ) ( x ) d η ~ ( w ˙ )
where
A 1 ( w ˙ ) = ∫ g 2 ( v ) ζ 1 ( v ) δ ( σ ( v ) ) 1 2 d μ 1 w ˙ ( v )
B 1 ( w ˙ ) = ∫ g 1 ( u ) ζ 2 ( u ) δ ( σ ( u ) ) 1 2 d μ 1 w ˙ ( u )
Moreover, if f   is of the form f ( x ) = g 1 ( r ( x ) ) g ( q ( x ) ) g 2 ( d ( x ) )   and ζ 1 , ζ 2 ∈ ∫ G ( 0 ) ⊕   ℋ ( e ( u ) ) d η 2 ( u )   , then
〈 V W L 1 ( f ) W * V * ζ 1 , ζ 2 〉
= ∫ ∫ g ( x ) δ ( x ) − 1 2 〈 L ( x ) A 2 ( w ˙ ) , B 2 ( w ˙ ) 〉 d β e ( u ) e ( u ) ( x ) d η ~ ( w ˙ )
= 〈 W L 2 ( Φ ( f ) ) W * ζ 1 , ζ 2 〉
where
A 2 ( w ˙ ) = ∫ g 2 ( v ) V * ζ 1 ( v ) δ ( σ ( v ) ) 1 2 d μ 1 w ˙ ( v )
= ∫ U v ˙ ( g 2 ) ( v ) ζ 1 ( v ) δ ( σ ( v ) ) 1 2 d μ 2 w ˙ ( v )
B 2 ( w ˙ ) = ∫ g 1 ( v ) V * ζ 2 ( v ) δ ( σ ( v ) ) 1 2 d μ 1 w ˙ ( v )
= ∫ U u ˙ ( g 1 ) ( u ) ζ 2 ( u ) δ ( σ ( u ) ) 1 2 d μ 2 w ˙ ( u )
Therefore ∥ L 1 ( f ) ∥ = ∥ L 2 ( Φ ( f ) ) ∥   . Consequently we can extend Φ   to a *   -homomorphism between the M σ * ( G , ν 1 )   and M σ * ( G , ν 2 )   . It is not hard to see that Φ   is in fact a *   -isomorphism: Φ − 1 ( f ) = ( x → U π ( r ( x ) ) * ( g 1 ) ( r ( x ) ) g ( q ( x ) ) U π ( d ( x ) ) * ( g 2 ) ( d ( x ) ) )   for each f   of the form f ( x ) = g 1 ( r ( x ) ) g ( q ( x ) ) g 2 ( d ( x ) ) .  

4 The case of locally transitive groupoids

A locally compact transitive groupoid G   is a groupoid for which all orbits [ u ]   are open in G ( 0 )   . We shall prove that if G   is a locally compact second countable locally transitive groupoid endowed with a Haar system ν   , then C * ( G , ν ) = M * ( G , ν ) = M σ * ( G , ν )   for any regular cross section σ .  
Notation 10 Let { ν u , u ∈ G ( 0 ) }   be a fixed Haar system on G   . Let μ   be a quasi-invariant measure, Δ   its modular function, ν 1   be the measure induced by μ   on G   and ν 0 = Δ − 1 2 ν 1   . Let us denote by I I μ ( G )   the set { f ∈ L 1 ( G , ν 0 ) : ∥ f ∥ I I , μ < ∞ } ,   where ∥ f ∥ I I , μ   is defined by ∥ f ∥ I I , μ = sup { ∫ | f ( x ) j ( d ( x ) ) k ( r ( x ) ) | d ν 0 ( x ) , ∫ | j | 2 d μ = ∫ | k | 2 d μ = 1 } .   If μ 1   and μ 2   are two equivalent quasi-invariant measures, then ∥ f ∥ I I , μ 1 = ∥ f ∥ I I , μ 2   , because ∥ f ∥ I I , μ = ∥ I I μ ( | f | ) ∥   for each quasi-invariant measure μ   , where I I μ   is the one dimensional trivial representation on μ   Define ∥ f ∥ I I   to be sup { ∥ f ∥ I I , μ : μ quasi-invariant Radon measure on G ( 0 ) }   The supremum can be taken over the classes of quasi-invariant measure.
If ∥ ∥   is the full C *   -norm on C c ( G )   , then ∥ f ∥ ≤ ∥ f ∥ I I for all f .   (see [7] )
Lemma 11 Let G   be a locally compact second countable groupoid with proper orbit space. Let { ν u , u ∈ G ( 0 ) }   be a Haar system on G   , ( { β v u } , { μ u ˙ } )   its decomposition over the principal groupoid associated to G   and δ   the associated 1   -cocycle. If f   is a universally measurable function on G   , then ∥ f ∥ I I ≤ sup w ˙ ( ∫ ∫ ( ∫ | f ( x ) | δ ( x ) − 1 2 d β v u ( x ) ) 2 d μ w ˙ ( v ) d μ w ˙ ( u ) ) 1 2 .  
Proof. Each Radon quasi-invariant measure is equivalent with a Radon measure of the form ∫ μ u ˙ d μ ~ ( u ˙ )   , where μ ~   is a probability measure on the orbit space G / G ( 0 ) .   Therefore for the computation of ∥ . ∥ I I   it is enough to consider only the quasi-invariant measures of the form μ = ∫ μ u ˙ d μ ~ ( u ˙ )   , where μ ~   is a probability measure on G ( 0 ) / G   . It is easy to see that the modular function of ∫ μ u ˙ d μ ~ ( u ˙ )   is Δ = δ   .
Let j , k ∈ L 2 ( G ( 0 ) , μ )   with ∫ | j | 2 d μ = ∫ | k | 2 d μ = 1   . We have
∫ ∫ ∫ ∫ | f ( x ) | δ ( x ) − 1 2 d β v u ( x ) | j ( v ) | | k ( u ) | d μ w ˙ ( v ) d μ w ˙ ( u ) d μ ~ ( w ˙ )
≤ ∫ ( ∫ ∫ ( ∫ | f ( x ) | δ ( x ) − 1 2 d β v u ( x ) ) 2 d μ w ˙ ( v ) d μ w ˙ ( u ) ) 1 2 ⋅
⋅ ( ∫ ∫ | j ( v ) | 2 | k ( u ) | 2 d μ w ˙ ( v ) d μ w ˙ ( u ) ) 1 2 d μ ~ ( w ˙ )
= sup w ˙ ( ∫ ∫ ( ∫ | f ( x ) | δ ( x ) − 1 2 d β v u ( x ) ) 2 d μ w ˙ ( v ) d μ w ˙ ( u ) ) 1 2 ⋅
⋅ ∫ ( ∫ | j ( v ) | 2 d μ w ˙ ( v ) ) 1 2 ( | k ( u ) | 2 d μ w ˙ ( u ) ) 1 2 d μ ~ ( w ˙ )
≤ sup w ˙ ( ∫ ∫ ( ∫ | f ( x ) | δ ( x ) − 1 2 d β v u ( x ) ) 2 d μ w ˙ ( v ) d μ w ˙ ( u ) ) 1 2
Consequently, ∥ f ∥ I I ≤ sup w ˙ ( ∫ ∫ ( ∫ | f ( x ) | δ ( x ) − 1 2 d β v u ( x ) ) 2 d μ w ˙ ( v ) d μ w ˙ ( u ) ) 1 2  
If G   is locally transitive, each orbit [ u ]   is open in G ( 0 )   . Each measure μ u ˙   is supported on [ u ]   . Since ( [ u ] )   is a partition of G ( 0 )   into open sets, it follows that there is a unique Radon measure m   on G ( 0 )   such that the restriction of m   at C c ( [ u ] )   is μ u ˙   for each [ u ]   .
Corollary 12 Let G   be a locally compact second countable locally transitive groupoid endowed with a Haar system ν = { ν u , u ∈ G ( 0 ) }   . Let f   be a universally measurable function such that ∥ f ∥ I I < ∞   .
  • 1. If ( f n ) n   is a uniformly bounded sequence of universally measurable functions supported on a compact set, and if ( f n ) n   converges pointwise to f   , then ( f n ) n   converges to f   in the norm of C * ( G , ν ) .  
  • 2. If ( f n ) n   is an increasing sequence of universally measurable nonnegative functions on G   that converges pointwise to f   , then ( f n ) n   converges to f   in the norm of C * ( G , ν ) .  
Proof. Let ( { β v u } , { μ u ˙ } )   be the decomposition of the Haar system over the principal groupoid associated to G   and δ   the associated 1   -cocycle. Let m   be the unique measure such that restriction of m   at C c ( [ u ] )   is μ u ˙   for each [ u ]   . Let ( f n ) n   be a sequence of universally measurable functions supported on a compact set K   . Let us write M = sup u , v β u v ( K − 1 ) .   and let us assume that ( f n ) n   converges pointwise to f   . According to Lemma  11 , we have ∥ f − f n ∥ I I ≤ sup w ˙ ( ∫ ∫ ( ∫ | f ( x ) − f n ( x ) | δ ( x ) − 1 2 d β v u ( x ) ) 2 d μ w ˙ ( v ) d μ w ˙ ( u ) ) 1 2 .   Hence
∥ f − f n ∥ I I ≤ sup w ˙ M ( ∫ ∫ ( ∫ | f ( x ) − f n ( x ) | 2 d β v u ( x ) ) d μ w ˙ ( v ) d μ w ˙ ( u ) ) 1 2
≤ M ( ∫ ∫ ( ∫ | f ( x ) − f n ( x ) | 2 d β u v ( x ) ) d m ( v ) d m ( u ) ) 1 2 .
If ∥ ∥   denotes the C *   -norm, then lim n ∥ f − f n ∥ ≤ lim n ∥ f − f n ∥ I I = 0 ,   because ∫ ∫ ( ∫ | f ( x ) − f n ( x ) | 2 d β u v ( x ) ) d m ( v ) d m ( u )   converges to zero, by the Dominated Convergence Theorem. Let ( f n ) n   be an increasing sequence of universally measurable nonnegative functions that converges pointwise to f .   Since
∥ f − f n ∥ I I ≤ sup w ˙ ( ∫ ∫ ( ∫ | f ( x ) − f n ( x ) | δ ( x ) − 1 2 d β v u ( x ) ) 2 d μ w ˙ ( v ) d μ w ˙ ( u ) ) 1 2
≤ ( ∫ ∫ ( ∫ | f ( x ) − f n ( x ) | δ ( x ) − 1 2 d β v u ( x ) ) 2 d m ( v ) d m ( u ) ) 1 2
it follows that lim n ∥ f − f n ∥ I I = 0 .  
Proposition 13 Let G   be a locally compact second countable locally transitive groupoid endowed with a Haar system ν = { ν u , u ∈ G ( 0 ) }   .Then any function in ℬ c ( G )   , the space of compactly supported Borel bounded function on G   , can be viewed as an element of C * ( G , ν )   .
Proof. Let ( { β v u } , { μ u ˙ } )   be the decomposition of the Haar system over the principal groupoid associated to G   , and δ   the associated 1   -cocycle. Let m   be the unique measure such that restriction of m   at C c ( [ u ] )   is μ u ˙   for each [ u ]   .
Let m   be a dominant for the family { μ u ˙ }   . Let ν 1   be the measure on G   define by ∫ f ( x ) d ν 1 ( x ) = ( ∫ ∫ ( ∫ f ( x ) d β u v ( x ) ) d m ( v ) d m ( u ) )   for all Borel nonnegative function f   . If f ∈   ℬ c ( G )   , then f   is the limit in L 2 ( G , ν 1 )   of a sequence, ( f n ) n   , in C c ( G )   that is supported on some compact set K   supporting f   . If we write M = sup u , v β u v ( K − 1 ) ,   we have
∥ f − f n ∥ I I ≤ sup w ˙ M ( ∫ ∫ ( ∫ | f ( x ) − f n ( x ) | 2 d β v u ( x ) ) d μ w ˙ ( v ) d μ w ˙ ( u ) ) 1 2
≤ M ( ∫ ∫ ( ∫ | f ( x ) − f n ( x ) | 2 d β u v ( x ) ) d m ( v ) d m ( u ) ) 1 2 .
If ∥ ∥   denotes the C *   -norm, then lim n ∥ f − f n ∥ ≤ lim n ∥ f − f n ∥ I I = 0 .   Thus f   can be viewed as an element in C * ( G , ν )   .
Proposition 14 If G   is a locally compact second countable locally transitive groupoid endowed with a Haar system { ν u , u ∈ G ( 0 ) }   with bounded decomposition, then C * ( G , ν ) = M * ( G , ν ) .  
Proof. It follows from the Proposition  13 .
Remark 15 Let G   be locally compact locally transitive groupoid. Let F   be a subset of G ( 0 )   containing only one element e ( u )   in each orbit [ u ]   . It is easy to see that F   is a closed subset of G   and that F   is a discrete space. Let σ : G ( 0 ) → G F   be a regular cross section of d F   . Let us endow ∪ [ u ] [ u ] × G e ( u ) e ( u ) × [ u ]   with the topology induced from G ( 0 ) × G F F × G ( 0 )   . The topology of ∪ [ u ] [ u ] × G e ( u ) e ( u ) × [ u ]   is locally compact because ∪ [ u ] [ u ] × G e ( u ) e ( u ) × [ u ]   is a closed subset of the locally compact space G ( 0 ) × G F F × G ( 0 )   . With the operations:
( u , x , v ) ( v , y , w ) = ( u , x y , w )
( u , x , v ) − 1 = ( v , x − 1 , u )
∪ [ u ] [ u ] × G e ( u ) e ( u ) × [ u ]   becomes a groupoid. Let us define φ : G → ∪ [ u ] [ u ] × G e ( u ) e ( u ) × [ u ]   by φ ( x ) = ( r ( x ) , σ ( r ( x ) ) x σ ( d ( x ) ) − 1 , d ( x ) )   and note that φ   is a Borel isomorphism which carries compact sets to relatively compact sets .
Lemma 16 Let G   be locally compact second countable locally transitive groupoid.
Let F   be a subset of G ( 0 )   containing only one element e ( u )   in each orbit [ u ]   . Let σ : G ( 0 ) → G F   be a regular cross section of d F   . Then any compactly supported Borel bounded function on G   is pointwise limit of a uniformly bounded sequence ( f n ) n   of Borel functions supported on a compact set supporting f   , having the property that each f n   is a linear combination of functions of the form x → g 1 ( r ( x ) ) g ( σ ( r ( x ) ) x σ ( d ( x ) ) − 1 ) g 2 ( d ( x ) )   where g 1 , g 2   are compactly supported Borel bounded function on G ( 0 )   and g   is a compactly supported Borel bounded function on G F F   .
Proof. Let us endow ∪ [ u ] [ u ] × G e ( u ) e ( u ) × [ u ]   with the topology induced from G ( 0 ) × G F F × G ( 0 )   as in Remark. 15 . The topology of ∪ [ u ] [ u ] × G e ( u ) e ( u ) × [ u ]   is locally compact. Any compactly supported Borel bounded function on G ( 0 ) × G F F × G ( 0 )   is pointwise limit of uniformly bounded sequences ( f n ) n   of Borel functions supported on a compact set, such that each function f n   is a linear combination of functions of the form ( u , x , v ) → g 1 ( u ) g ( x ) g 2 ( v )   where g 1 , g 2   are compactly supported Borel bounded function on G ( 0 )   and g   is a compactly supported Borel bounded function on G F F   . Consequently, any compactly supported Borel bounded function on ∪ [ u ] [ u ] × G e ( u ) e ( u ) × [ u ]   has the same property. Since φ : G → ∪ [ u ] [ u ] × G e ( u ) e ( u ) × [ u ]   defined by φ ( x ) = ( r ( x ) , σ ( r ( x ) ) x σ ( d ( x ) ) − 1 , d ( x ) )   is a Borel isomorphism which carries compact sets to relatively compact sets, it follows that any compactly supported Borel bounded function on G   can be represented as a pointwise limit of a uniformly bounded sequence ( f n ) n   of Borel functions supported on a compact set supporting f   , having the property that each f n   is a linear combination of functions of the form x → g 1 ( r ( x ) ) g ( σ ( r ( x ) ) x σ ( d ( x ) ) − 1 ) g 2 ( d ( x ) )  
Corollary 17 Let G   be locally compact second countable locally transitive groupoid. Let F   be a subset of G ( 0 )   containing only one element e ( u )   in each orbit [ u ]   . Let σ : G ( 0 ) → G F   be a regular cross section of d F   . Then the linear span of the functions of the form x → g 1 ( r ( x ) ) g ( σ ( r ( x ) ) x σ ( d ( x ) ) − 1 ) g 2 ( d ( x ) )   where g 1 , g 2 ∈ ℬ c ( G ( 0 ) )   and g ∈ ℬ c ( G F F ) ,   is dense in the full C *   -algebra of G   .
Proof. Let f   be a function on G   , defined by f ( x ) = g 1 ( r ( x ) ) g ( σ ( r ( x ) ) x σ ( d ( x ) ) − 1 ) g 2 ( d ( x ) )   where g 1 , g 2 ∈ ℬ c ( G ( 0 ) )   and g ∈ ℬ c ( G F F )   . Then f ∈ ℬ c ( G )   , therefore it can be viewed as an element of the C * ( G , ν )   as we note in Proposition  13 . Each f ∈ ℬ c ( G )   (in particular in C c ( G )   ) is the limit (pointwise and consequently in the C *   -norm according to Corollary  12 ) of a uniformly bounded sequence ( f n ) n   of Borel functions supported on a compact set supporting f   , having the property that each f n   is a linear combination of functions of the required form.
Proposition 18 Let G   is a locally compact second countable locally transitive groupoid endowed with a Haar system { ν u , u ∈ G ( 0 ) }   . Let F   be a subset of G ( 0 )   containing only one element e ( u )   in each orbit [ u ]   . Let σ : G ( 0 ) → G F   be a regular cross section of d F   . Then C * ( G , ν ) = M * ( G , ν ) = M σ * ( G , ν ) .  
Proof. We have proved that C * ( G , ν ) = M * ( G , ν )   . From the preceding corollary, it follows that the linear span of the functions of the form x → g 1 ( r ( x ) ) g ( σ ( r ( x ) ) x σ ( d ( x ) ) − 1 ) g 2 ( d ( x ) )   where g 1 , g 2 ∈ ℬ c ( G ( 0 ) )   and g ∈ ℬ c ( G F F )   is dense in C * ( G , ν )   . But this space is contained in ℬ σ ( G )   . Therefore C * ( G , ν ) = M * ( G , ν ) = M σ * ( G , ν )   .

5 The case of principal proper groupoids case

Notation 19 Let G   be a locally compact second countable groupoid with proper orbit space. Let F   be a Borel subset of G ( 0 )   containing only one element e ( u )   in each orbit [ u ]   . Let σ : G ( 0 ) → G F   be a regular cross section for d F : G F → G ( 0 ) , d F ( x ) = d ( x )   with σ ( e ( v ) ) = e ( v )   for all v ∈ G ( 0 )   . Let q : G → G F F   be defined by q ( x ) = σ ( r ( x ) ) x σ ( d ( x ) ) − 1   We shall endow G F F   with the quotient topology induced by q .   We shall denote by C σ ( G )   the linear span of the functions of the form x → g 1 ( r ( x ) ) g ( σ ( r ( x ) ) x σ ( d ( x ) ) − 1 ) g 2 ( d ( x ) )   where g 1 , g 2 ∈ C c ( G ( 0 ) )   and g ∈ C c ( G F F )   .
Proposition 20 With the Notation  19 , if the space of continuous functions with compact support on G F F   (with the respect to the quotient topology induced by q   ) separates the points of G F F   , then C σ ( G )   is dense in C c ( G )   (for the inductive limit topology). In particular, if the quotient topology induced by q   on G F F   is a locally compact (Hausdorff ) topology, then C σ ( G )   is dense in C c ( G )   .
Proof. If the space of continuous functions with compact support on G F F   (with the respect to the quotient topology induced by q   ) separates the points of G F F   , then C σ ( G )   separates the points of G   . By Stone-Weierstrass Theorem, it follows that C σ ( G )   is dense in C c ( G )   (for the inductive limit topology)
Proposition 21 Let G   be a locally compact principal groupoid. If G   is proper, then the quotient topology induced by q   on G F F   is a locally compact (Hausdorff ) topology. Consequently, C σ ( G )   is dense in C c ( G )   for the inductive limit topology (we use the Notation  19 ).
Proof. Let π : G → G ( 0 ) / G   be the canonical projection. Let us note that for a principal groupoid the condition q ( x ) = q ( y )   is equivalent with π ( r ( x ) ) = π ( r ( y ) ) .   First we shall prove that the topology on G F F   is Hausdorff. Let ( x i ) i   and ( y i ) i   be two nets with q ( x i ) = q ( y i )   for every i   . Let us suppose that ( x i ) i   converges to x   and ( y i ) i   converges to y   . Then lim π ( r ( x i ) ) = lim π ( r ( y i ) ) = π ( r ( x ) ) = π ( r ( y ) )   Hence q ( x ) = q ( y )   , and therefore the topology on G F F   is Hausdorff. We shall prove that q   is open. If ( z i ) i   is a net converging to q ( x )   in G F F   , then π ∘ r ( z i )   converges to π ∘ r ( x )   . Since π ∘ r : G → G ( 0 ) / G   is an open map, there is a net ( x i ) i   converging to x   , such that π ∘ r ( x i ) = π ∘ r ( z i )   , and consequently q ( x i ) = q ( z i ) = z i   . Hence q   is an open map and the quotient topology induced by q   on G F F   is locally compact.
Theorem 22 Let G   be a locally compact second countable groupoid with proper orbit space. Let F   be a Borel subset of G ( 0 )   meeting each orbit exactly once. Let σ : G ( 0 ) → G F   be a regular cross section for d : G F → G   . Let us assume that the quotient topology induced by q   on G F F   is a locally compact (Hausdorff ) topology. Let { ν u , u ∈ G ( 0 ) }   be a Haar system on G   . Then C * ( G , ν ) ⊂ M σ * ( G , ν ) ⊂ M * ( G , ν ) .  
Proof. From Proposition  20  C σ ( G )   is dense in C c ( G )   for the inductive limit topology and hence is dense in C * ( G , ν )   . Since C σ ( G ) ⊂ ℬ σ ( G )   , it follows that C * ( G , ν ) ⊂ M σ * ( G , ν )   .
Corollary 23 Let G   be a locally compact second countable principal proper groupoid. Let F   be a Borel subset of G ( 0 )   meeting each orbit exactly once. Let σ : G ( 0 ) → G F   be a regular cross section for d : G F → G   . Let { ν u , u ∈ G ( 0 ) }   be a Haar system on G   . Then C * ( G , ν ) ⊂ M σ * ( G , ν ) ⊂ M * ( G , ν ) .  
Proof. Applying Proposition  21 , we obtain that the quotient topology induced by q   on G F F   is a locally compact (Hausdorff ) topology. Therefore G   satisfies the hypothesis of Theorem  22 .
Definition 24 Let { μ u ˙ } u ˙   be a system of measures on G ( 0 )   satisfying:
  • 1. s u p p ( μ u ˙ ) = [ u ]   for all u ˙   .
  • 2. For all compactly supported continuous functions f   on G ( 0 )   the function u → ∫ f ( v ) μ π ( u ) ( v )   is continuous
We shall say that the Hilbert bundle determined by the system of measures { μ u ˙ } u ˙   has a continuous basis if there is sequence f 1 , f 2 , . . .   of real valued continuous functions on G ( 0 )   such that dim ( L 2 ( μ u ˙ ) ) = ∞   if and only if ∥ f n ∥ 2   for n = 1 , 2 , . . .   and then { f 1 , f 2 , . . . }   gives an ortonormate basis of L 2 ( μ u ˙ )   , while dim ( L 2 ( μ u ˙ ) ) = k < ∞   if and only if ∥ f n ∥ 2 = 1   for n ≤ k   , and ∥ f n ∥ 2 = 0   for n > k   and then { f 1 , f 2 , . . . f k }   gives an ortonormate basis of L 2 ( μ u ˙ )   .
Remark 25 Let { μ 1 u ˙ } u ˙   and { μ 2 u ˙ } u ˙   be two systems of measures on G ( 0 )   satisfying:
  • 1. s u p p ( μ i u ˙ ) = [ u ]   for all u ˙   , i = 1 , 2   .
  • 2. For all compactly supported continuous functions f   on G ( 0 )   the function u → ∫ f ( v ) μ i π ( u ) ( v )   is continuous
Let us assume that the Hilbert bundles determined by the systems of measures { μ i u ˙ } u ˙   have continuous bases. Let f 1 . f 2 , . . .   be a continuous basis for Hilbert bundle determined by { μ 1 u ˙ } u ˙   and let g 1 , g 2 , . . .   be a continuous basis for Hilbert bundle determined by { μ 2 u ˙ } u ˙   . Let us define a unitary operator U u ˙ : L 2 ( μ 1 u ˙ ) → L 2 ( μ 2 u ˙ )   by U u ˙ ( f n ) = g n for all n   Then the family { U u ˙ } u ˙   has the following properties:
  • 1. For all Borel bounded function f   on G ( 0 )   , u → U π ( u ) ( f )   is a bounded Borel function with compact support.
  • 2. For all Borel bounded function f   on G ( 0 )   , U π ( u ) ( f ¯ ) = U π ( u ) ( f ) ¯  
  • 3. For all compactly supported continuous functions f   on G ( 0 )   there is a sequence ( h n ) n   of compactly supported continuous functions on G ( 0 )   such that s u p u ˙ ∫ | U u ˙ ( f ) − h n | 2 d μ 2 u ˙ → 0 ( n → ∞ )  
Indeed, we can define h n ( v ) = ∑ k = 1 n g k ( v ) ∫ f ( u ) f k ( u ) d μ 1 π ( v ) ( u ) .  
Remark 26 Let G   be a locally compact second countable groupoid with proper orbit space. Let F   be a Borel subset of G ( 0 )   containing only one element e ( u )   in each orbit [ u ]   . Let us assume that F ∩ [ K ]   has a compact closure for each compact subset K   of G ( 0 )   , and let σ : G ( 0 ) → G F   be a regular cross section for d F : G F → G ( 0 )   . Let us endow G F F   with the quotient topology induced by q : G → G F F   q ( x ) = σ ( r ( x ) ) x σ ( d ( x ) ) − 1 , x ∈ G   If g ∈ C c ( G F F )   and g 1 , g 2   are two functions on G ( 0 )   with the property that there is two sequences ( h n 1 ) n   and ( h n 2 ) n   of compactly supported continuous functions on G ( 0 )   such that s u p u ˙ ∫ | g i − h n i | 2 d μ 2 u ˙ → 0 ( n → ∞ )   for i = 1 , 2   , then x → f g 1 ( r ( x ) ) g ( σ ( r ( x ) ) x σ ( d ( x ) ) − 1 ) g 2 ( d ( x ) )   can be viewed as an element of C * ( G , ν )   . Indeed, it is easy to see that ‖ f − ( h n 1 ∘ r ) ( g ∘ q ) ( h n 2 ∘ d ) ‖ I I → 0 ( n → ∞ ) .  
Proposition 27 Let G   be a locally compact second countable principal proper groupoid. Let ν i = { ν i u , u ∈ G ( 0 ) } ,   i = 1 , 2   be two Haar system on G   and ( { β v u } , { μ i u ˙ } )   the corresponding decompositions over the principal groupoid. If the Hilbert bundles determined by the systems of measures { μ i u ˙ } u ˙   have continuous bases, then the C *   -algebras C * ( G , ν 1 )   and C * ( G , ν 2 )   are *   -isomorphic.
Proof. We use the Notation  19 . From Proposition  20  C σ ( G )   is dense in C c ( G )   for the inductive limit topology and hence is dense in C * ( G , ν 1 )   . We shall define a *   -homomorphism Φ   from C σ ( G )   to C * ( G , ν 2 )   . It suffices to define Φ   on the set of function on G   of the form x → g 1 ( r ( x ) ) g ( q ( x ) ) g 2 ( d ( x ) )   where g 1 , g 2 ∈ C c ( G ( 0 ) )   and g ∈ C c ( G F F )   . Let { U u ˙ } u ˙   be the family of unitary operators with the properties stated in Remark  25 associated to the systems of measures { μ i u ˙ } u ˙   , i = 1 , 2   .
Let us define Φ   by Φ ( f ) = ( x → U π ( r ( x ) ) ( g 1 ) ( r ( x ) ) g ( q ( x ) ) U π ( d ( x ) ) ( g 2 ) ( d ( x ) ) )   where f   is defined by f ( x ) = g 1 ( r ( x ) ) g ( q ( x ) ) g 2 ( d ( x ) )   with g 1 , g 2 ∈ C c ( G ( 0 ) )   and g ∈ C c ( G F F )   .
We noted in Remark  26 that the functions of the form Φ ( f )   can be viewed as elements of C * ( G , ν 2 )   . With the same argument as in the proof of Theorem  9 , it follows that Φ   can be extended to *   -isomorphism between C * ( G , ν 1 )   and C * ( G , ν 2 )   .
References

  1. C. Anantharaman-Delaroche, J. Renault, Amenable groupoids, Monographie de L'Enseignement Mathematique No 36, Geneve, 2000.
  2. G. Mackey, Induced representations of locally compact groups. I   , Ann. of Math., 55(1952), 101-139.
  3. P. Muhly, J. Renault and D. Williams, Equivalence and isomorphism for groupoid C*-algebras, J. Operator Theory 17(1987), 3-22.
  4. P. Muhly, Coordinates in operator algebra, (Book in preparation).
  5. A. Ramsay, Virtual groups and groups actions, Adv. in Math. 6(1971), 253-322.
  6. A. Ramsay, Topologies on measured groupoids, J. Funct. Anal. 47(1982), 314-343.
  7. A. Ramsay and M. E. Walter, Fourier-Stieltjes Algebras of locally compact groupoids, J. Funct. Anal. 148(1997), 314-367.
  8. J. Renault, A groupoid approach to C *   algebras, Lecture Notes in Math., Springer-Verlag, 793, 1980.
  9. J. Renault,The ideal structure of groupoid crossed product algebras, J. Operator Theory, 25(1991), 3-36.
  10. J. Westman, Nontransitive groupoid algebras, Univ. of California at Irvine, 1967.

University Constantin Brâncuşi of Târgu-Jiu Bulevardul Republicii, Nr. 1, 210152 Târgu-Jiu , Gorj Romania e-mail: ada@utgjiu.ro