November 27, 2006

2000 Mathematics Subject Classification. Primary 46L65, Secondary 46L08.Partially supported by Dinacyt (Proyecto Clemente Estable 8013), Uruguay.
<ph f="cmbx">Morita equivalence for Quantum Heisenberg Manifolds</ph>

Beatriz Abadie

Centro de Matematicas. Facultad de Ciencias. Igua 4225, CP 11 400, Montevideo, Uruguay. E-mail address : abadie@cmat.edu.uy
Introduction. Quantum Heisenberg manifolds { D c μ ν : c Z , c > 0 , μ , ν R }   were constructed by Rieffel in [Rf4as a quantization deformation of certain homogeneous spaces H / N c   , H   being the Heisenberg group.
It was shown in [Ab1,3.4that K 0 ( D c μ ν ) = Z 3 Z c   , which implies that D c μ ν   and D μ ν c   are not isomorphic unless c = c   . Besides, D c μ ν   and D μ ν c   are isomorphic when ( 2 μ , 2 ν )   and ( 2 μ , 2 ν )   belong to the same orbit under the usual action of GL 2 ( Z )   on T 2   ( [AE,Theorem2.2; see also [Ab2,3.3).
The range of traces on D c μ ν   was discussed in [Ab2, where it was shown that the range of the homomorphism induced on K 0 ( D c μ ν )   by any tracial state on D c μ ν   has range G μ ν = Z + 2 μ Z + 2 ν Z   . As a consequence ([Ab2,3.17), the isomorphism condition stated above turns out to be necessary when the rank of G μ ν   is either 1 or 3. Rieffel showed in [Rf4that D c μ ν   is simple if and only if { 1 , μ , ν }   is linearly independent over the field of rational numbers (i.e rank G μ ν = 3   ); it might be interesting to know whether in this case the classification can be made by means of the results of Elliott and Gong ([EG).
The quantum Heisenberg manifold D c μ ν   was described in [AEEas a crossed product by a Hilbert C *   -bimodule. In order to discuss Morita equivalence within this family, we adapt to this setting some of the techniques employed in the analogous discussion for non-commutative tori ([Rf3) and Heisenberg C *   -algebras ([Pa2). Thus we generalize in Section 1 Green's result (discussed by Rieffel in [Rf2,Situation10) on the Morita equivalence of the crossed products C 0 ( M / K ) H   and C 0 ( M / H ) K   , for free and proper commuting actions on a locally compact space M   . This result provides the main tool to discuss, in Section 2, Morita equivalence for quantum Heisenberg manifolds.

1 Morita equivalence of crossed products by certain Hilbert C *   -bimodules over commutative C *   -algebras.

For a Hilbert C *   -bimodule X   over a C *   -algebra A   , the crossed product A X   was introduced in [AEE(see also [Pi) as the universal C *   -algebra for which there exist a *-homomorphism i A : A A X   and a continuous linear map i X : X A X   such that
i X ( a x ) = i A ( a ) i X ( x ) , i A ( x , y L ) = i X ( x ) i X ( y ) * .
i X ( x a ) = i X ( x ) i A ( a ) , i A ( x , y R ) = i X ( x ) * i X ( y ) .
The crossed product A X   carries a dual action δ   of S 1   , defined by δ z ( i A ( a ) ) = i A ( a )   , δ z ( i X ( x ) ) = z i X ( x )   , for a A   , x X   and z S 1   .
Moreover, if a C *   -algebra B   carries an action δ   of S 1   such that B   is generated as a C *   -algebra by the fixed point subalgebra B 0 = { b B : δ z ( b ) = b z S 1 }   and the first spectral subspace B 1 = { b B : δ z ( b ) = z b z S 1 }   , then B   is isomorphic to B 0 B 1   (where B 1   has the obvious Hilbert C *   -bimodule structure over B 0   ), via an isomorphism that takes the action δ   into the dual action.
If X   is an A   -Hilbert C *   -bimodule and α Aut ( A )   , we denote by X α   the Hilbert C *   -bimodule over A   obtained by leaving unchanged the left structure, and by setting x X α a : = x α ( a ) , x , y R X α : = α 1 ( x , y R ) ,   where the undecorated notation refers to the original right structure of X   .
For α Aut ( A )   and the usual A   -Hilbert C *   -bimodule structure on A   , the crossed product A A α   is easily checked to be the usual crossed product A α Z   .
Definition 1.1. Given a proper action α   of Z   on a locally compact Hausdorff space M   and a unitary u C b ( M )   , let X α , u   denote the set of functions f C b ( M )   satisfying f = u α ( f )   , and such that the map x | f ( x ) |   , which is constant on α   -orbits, belongs to C 0 ( M / α )   .
Then X α , u   is a Hilbert C *   -bimodule over C 0 ( M / α )   for pointwise multiplication on the left and the right, and inner products given by f , g L = f g ¯   , f , g R = f ¯ g   .
Proposition 1.2. Let α   and β   be free and proper commuting actions of Z   on a locally compact Hausdorff space M   , and let u   be a unitary in C b ( M )   . Then the C *   -algebras C 0 ( M / α ) X α , u β   and C 0 ( M / β ) X β , u * α   are Morita equivalent.
  • Proof. Let U : Z × Z U ( C b ( M ) )   be given by U ( n , k ) = { 1 if either n = 0 or k = 0 . i S k , j S n α i β j ( u ) , for n , k > 0 i S k , j S n α i β j ( u * ) , for either n or k < 0 , and n k 0 ,   where S l = { 0 , 1 , . . . l 1 }   if l > 0   and S l = { 1 , 2 , . . . l }   if l < 0   .
    Straightforward computations show that U ( m + n , k ) = U ( m , k ) β m ( U ( n , k ) )   , and U ( n , k + l ) = U ( n , k ) α k ( U ( n , l ) )   . Consider the proper actions γ α   and γ β   of Z   on C 0 ( M ) β Z   and C 0 ( M ) α Z   , respectively, given by: [ γ k α ( φ ) ] ( n ) = U ( n , k ) α k [ φ ( n ) ] and [ γ n β ( ψ ) ] ( k ) = U * ( n , k ) β n [ ψ ( k ) ] ,   for φ C c ( Z , C 0 ( M ) ) C 0 ( M ) β Z   and ψ C c ( Z , C 0 ( M ) ) C 0 ( M ) α Z   .
    These two actions correspond, respectively, to γ α , U   and γ β , U *   in [Ab1,Propositions1.2and2.1. By virtue of [Ab1,Theorem2.12, the generalized fixed-point algebras, in the sense of [Rf5,Definition1.4, D α   and D β   of C 0 ( M ) β Z   and C 0 ( M ) α Z   under the actions γ α   and γ β   , respectively, are Morita equivalent. The result will then be proved once we show that D α = C 0 ( M / α ) X α , u β   and D β = C 0 ( M / β ) X β , u * α   .
    Recall from [Ab1,Proposition2.1that D α   is defined to be the closed span in ( C 0 ( M ) β Z )   of the set { P α ( φ * * ψ ) : φ , ψ C c ( Z × M ) }   , where P α ( φ ) ( x , n ) = k Z [ γ k α ( φ ) ] ( x , n ) ,   for φ C c ( Z × M ) C 0 ( M ) β Z , x M   , and n Z   .
    The C *   -algebra D α   can also be described ([Ab1,Proposition2.8) as the closure in ( C 0 ( M ) β Z )   of the *   -subalgebra C α = { F C c ( Z , C b ( M ) ) : γ α ( F ) = F   and π α ( supp F ( n ) )   is precompact for all n Z }   , where π α   denotes the canonical projection π α : M M / α   .
    Now, since C α   is contained in C b ( M ) β Z   , which is closed in ( C 0 ( M ) β Z )   , so is D α   . Moreover, the C *   -algebra D α   is invariant under the dual action β ^   of T   on C b ( M ) β Z   :
    [ γ α ( β ^ z F ) ] ( n , x ) = U ( n , 1 ) ( x ) ( β ^ z ( F ) ) ( n , α 1 x )
    = U ( n , 1 ) ( x ) z n F ( n , α 1 x )
    = z n F ( n , x )
    = ( β ^ z F ) ( n , x ) ,
    for F C α   , x M   , n Z   , and z T .   Besides, supp ( β ^ z ( F ) ( n ) ) = supp F ( n )   for all n Z   , so β ^ z ( F ) C α   for all z T   .
    We next show that the action β ^   on D α   is semi-saturated. That is, that, as a C *   -algebra, D α   is generated by the fixed-point subalgebra D 0   and the first spectral subspace D 1 = { d D α : β ^ z ( d ) = z d z T }   for the restriction of the dual action β ^   .
    Since the maps P i : D α D i   given by P i ( a ) = T z i β ^ z ( a ) d z   are surjective contractions, D i   is the closure of P i ( C α )   . Now, for i = 0 , 1   , C i = C α F i   , and D i = D α F i   , where F 0   and F 1   are, respectively, the fixed-point subalgebra and the first spectral subspace of C b ( M ) β Z   , which are known to be the δ i   -maps; that is, F i = { F C c ( Z , C b ( M ) ) : supp F = { i } }   .
    Notice that C α F 0 = { f δ 0 : f C b ( M ) : π α ( supp f ) is precompact and f = α ( f ) }   can be identified with C c ( M / α )   via f δ 0 f ~   , where f ~ π α = f   , and that this map extends to a *   -isomorphism between D 0   and C 0 ( M / α )   . Now, D 1   is a Hilbert C *   -bimodule over D 0   for
    (1b) (1a) ( f δ 0 ) * ( g δ 1 ) = ( f g ) δ 1 and ( g δ 1 ) * ( f δ 0 ) = ( g β ( f ) ) δ 1 , f δ 1 , g δ 1 L = ( f δ 1 ) * ( g δ 1 ) * = ( f g ¯ ) δ 0 , f δ 1 , g δ 1 R = ( f δ 1 ) * * ( g δ 1 ) = ( β 1 ( f ¯ g ) ) δ 0 . (1c)
    Notice that D 1   is full on the left (and on the right, by a similar argument) as a Hilbert C *   -bimodule over C 0 ( M / α )   . For < D 1 , D 1 > L   , the closed linear span in C 0 ( M / α ) = D 0   of the set { f δ 1 , g δ 1 L : f δ 1 , g δ 1 D 1 }   , is a closed ideal of C 0 ( M / α )   . Therefore, unless < D 1 , D 1 > L = C 0 ( M / α )   , there exists x 0 M   such that f ( x 0 ) = 0   for all f δ 1 D 1   .
    Now, given x 0 M   , we can choose ([Rf2,Situation10) a neighborhood U   of x 0   such that U α k ( U ) =   for k 0   . Let g C c ( M ) +   be such that supp g U   and g ( x 0 ) = 1   .
    Then
    [ P α ( ( g 1 / 2 δ 0 ) * * ( g 1 / 2 δ 1 ) ) ] ( x , n ) = = ( P α ( g δ 1 ) ) ( x , n ) = ( k U ( 1 , k ) ( x ) g ( α k ( x ) ) ) δ 1 ( n ) ,  
    so P α ( ( g 1 / 2 δ 0 ) * * ( g 1 / 2 δ 1 ) ) D 1   , and equals 1 at ( x 0 , 1 )   .
    In order to prove that C α C * ( D 0 , D 1 )   , it suffices to show that f δ k C * ( D 0 , D 1 )   for f δ k C α   , k Z   ; since C α   is closed under involution, we may assume that k 0   . We show this fact, which clearly holds for k = 0   and k = 1   , by induction on k   .
    If f δ k C α   and ε > 0   , since π α ( supp f )   is precompact in M / α   , and D 1   is full over C 0 ( M / α )   , we can find φ i , ψ i D 1 , i = 1 , . . . , p ,   such that i ( φ i * ψ i * ) * f δ 1 f δ 1 D α = i φ i , ψ i L f f C b ( M ) < ε   .
    Now, since φ i   and ψ i * * f   belong to C * ( D 0 , D 1 )   for i = 1 , . . . , p   , so does f   . This shows that D α = C * ( D 0 , D 1 )   and, consequently, by [AEE,Theorem3.1, that D α = D 0 D 1   .
    It only remains to notice now that D 0 D 1 = C 0 ( M / α ) X β α , u   . As noticed above, D 0   is isomorphic to C 0 ( M / α )   . On the other hand, the map f δ 1 f   takes C α F 1   to X β α , u   . By keeping track of the formulae in  1 , one easily checks that that map is an isometry, so it extends to an isometry from D 1   to X β α , u   , which is onto because its image contains the dense set: X 0 α , u = { f X β α , u : the map x | f ( x ) | is compactly supported on M / α } .   (Notice that X 0 α , u   is dense in X β α , u   , because, if { e λ }   is an approximate identity for C c ( M / α )   , then e λ f   converges to f   for all f X β α , u   ). This shows that D α   is isomorphic to C 0 ( M / α ) X β α , u   . Analogously, D β   is isomorphic to C 0 ( M / β ) X α β , u *   .

2 Morita equivalence for quantum Heisenberg manifolds

In [AEE(see also [AE,2) the quantum Heisenberg manifold D c μ ν   was shown to be the crossed product of C ( T 2 )   , the C *   -algebra of continuous functions on the torus, by the Hilbert C *   -bimodule M α μ ν c   , where α μ ν ( x , y ) = ( x + 2 μ , y + 2 ν )   , and M c = { f C b ( R × T ) : f ( x + 1 , y ) = e 2 π i c y f ( x , y ) }   is the Hilbert C *   -bimodule obtained by letting C ( T 2 )   act by pointwise product, and by defining the inner products f , g L = f g ¯   , f , g R = f ¯ g   .
Remark 2.1. The C *   -algebras D c μ ν   and D c μ ν   are isomorphic when the projections of ( 2 μ , 2 ν )   and ( 2 μ , 2 ν )   on the torus are in the same orbit under the usual action of GL 2 ( Z )   ([AE,Theorem2.2, see also [Ab2,Remark3.3).
Proposition 2.2. Let μ 0   . Then D c μ ν   and D 1 4 μ , ν 2 μ c   are Morita equivalent.
  • Proof. We follow the lines of [Rf3,1.1and apply Proposition  1.2 to the following setting: α   and β   consist of translation on R × T   by ( 1 2 μ , 0 )   and ( 1 , 2 ν )   , respectively, and u C b ( R × T )   is given by u ( x , y ) = e ( c y )   , where T   is viewed as R / Z   and, for a real number h   , e ( h ) = e 2 π i h   .
    Then, by Proposition  1.2 , C ( ( R × T ) / α ) X β α , u   and C ( ( R × T ) / β ) X α β , u *   are Morita equivalent, where
    X α , u = { F C b ( R × T ) : F ( x 1 2 μ , y ) = e ( c y ) F ( x , y ) } and X β , u * = { F C b ( R × T ) : F ( x 1 , y 2 ν ) = e ( c y ) F ( x , y ) } = { F C b ( R × T ) : F ( x + 1 , y + 2 ν ) = e ( c ( y + 2 ν ) ) F ( x , y ) } .  
    Let H α : C ( T 2 ) C ( ( R × T ) / α )   and H β : C ( T 2 ) C ( ( R × T ) / β )   be the isomorphisms given by: ( H α φ ) ( x , y ) = φ ( 2 μ x , y ) , ( H β φ ) ( x , y ) = φ ( x , 2 ν x y ) ,   and, for ( μ , ν ) = ( 1 4 μ , ν 2 μ )   , set J α : M α μ ν c X β α , u and J β : M α μ ν c X α β , u * ,   ( J α f ) ( x , y ) = f ( 2 μ x , y ) , ( J β f ) ( x , y ) = e ( c x ( x + 1 ) ν ) f ( x , 2 ν x y ) .   Notice that ( J α f ) ( x 1 2 μ , y ) = f ( 2 μ x 1 , y ) = e ( c y ) ( J α f ) ( x , y ) ,   and
    ( J β f ) ( x + 1 , y + 2 ν ) = e ( c ( x + 1 ) ( x + 2 ) ν ) f ( x + 1 , 2 ν x y )
    = e ( c ( x + 1 ) ( x + 2 ) ν ) e ( c ( 2 ν x y ) ) f ( x , 2 ν x y )
    = e ( c ( y + 2 ν ) ) ( J β f ) ( x , y ) ,
    so the definitions make sense.
    For i = α , β   , it is easily checked that J i   is a bijection and that, for φ C ( T 2 )   , f , g M c   : J i ( φ f ) = H i ( φ ) J i ( f ) , J i ( f φ ) = J i ( f ) H i ( φ ) ,   J i f , J i g L = H i ( f , g L ) , J i f , J i g R = H i ( f , g R ) .   This shows that D c μ ν = C ( T 2 ) M α μ ν c   and D c μ ν = C ( T 2 ) M α μ ν c   are isomorphic, respectively, to C ( ( R × T ) / α ) X β α , u   and C ( ( R × T ) / β ) X α β , u *   , and they are, consequently, Morita equivalent to each other.
Corollary 2.3. Let μ Q   , and let A = ( a b c d ) GL 2 ( Z )   . If 2 μ = 2 a μ + b 2 c μ + d and 2 ν = 2 ν 2 c μ + d ,   then the quantum Heisenberg manifolds D c μ ν   and D c μ ν   are Morita equivalent.
  • Proof. It suffices to check the statement for A 1 = ( 1 1 0 1 )   and A 2 = ( 0 1 1 0 )   , since A 1   and A 2   generate GL 2 ( Z )   ([Ku,AppendixB), and ( μ , ν ) ( μ , ν )   defines an action of GL 2 ( Z )   on ( R \ Q ) × R   . For A = A 1   we get isomorphic C *   -algebras by Remark  2.1 . For A = A 2   , we get ( μ , ν ) = ( 1 4 μ , ν 2 μ )   , and the result follows from Proposition  2.2 .
Proposition 2.4. Let { 1 , μ , ν }   be linearly independent over Q   , and let A = ( a b c d e f g h i ) GL 3 ( Z )   . If 2 μ = 2 a μ + 2 b ν + c 2 g μ + 2 h ν + i and 2 ν = 2 d μ + 2 e ν + f 2 g μ + 2 h ν + i ,   then the quantum Heisenberg manifolds D c μ ν   and D c μ ν   are Morita equivalent.
  • Proof. As in the proof of Theorem 1.7 in [Pa2, A = A 1 A 2 A 3   , where A 1 = ( A B C D E F 0 0 1 ) , A 2 ( G 0 H 0 1 0 I 0 J ) , A 3 = ( K L 0 M N 0 0 0 1 ) ,   and A i GL 3 ( Z )   , for i = 1 , 2 , 3   .
    Since the map ( μ , ν ) ( μ , ν )   defines an action of GL 3 ( Z )   on the set { ( μ , ν ) R 2 : { 1 , μ , ν } is linearly independent over Q }   , it suffices to check the statement for A i   , i = 1 , 2 , 3 .   For A = A 1   and A = A 3   the C *   -algebras D c μ ν   and D c μ ν   are isomorphic by Remark  2.1 . Thus it suffices to show the result for A = A 2   . The map ( G H I J ) ( G 0 H 0 1 0 I 0 J )   is a group homomorphism from GL 2 ( Z )   into GL 3 ( Z )   , and GL 2 ( Z )   is generated by ( 1 1 0 1 )   and ( 0 1 1 0 )   , which implies that we only need to prove the statement for A 1 = ( 1 0 1 0 1 0 0 0 1 )   and A 2 = ( 0 0 1 0 1 0 1 0 0 )   . For A 1   we get 2 μ = 2 μ + 1   , 2 ν = 2 ν   , so D c μ ν   and D c μ ν   are isomorphic by Remark  2.1 . Proposition  2.2 takes care of the case A = A 2   , since then we have ( μ , ν ) = ( 1 4 μ , ν 2 μ )   .
Notation 2.5. We denote by G μ ν   the subgroup of R   generated by { 1 , 2 μ , 2 ν }   . It was shown in [Ab2,Theorem3.16that the homomorphism induced on K 0 ( D c μ ν )   by any tracial state on D c μ ν   has range G μ ν   .
Remark 2.6. If rank G μ ν = 2   , then there exist an irrational number ν   and integers p , q Z , p 0   , ( p , q ) = 1   , such that D c μ ν   and D p 2 q , ν c   are isomorphic.
  • Proof. We proceed as in [Pa1,Proposition1.5. Let μ 0 = 2 μ   , ν 0 = 2 ν   . Since the group generated by { 1 , μ 0 , ν 0 }   has rank 2, either μ 0   or ν 0   is irrational.
    We may assume that ν 0   is irrational, because, by Remark  2.1 , D c μ ν   and D ν μ c   are isomorphic. Besides, there exist M , N , P Z   , with N 0   such that M + N μ 0 + P ν 0 = 0   , so we have μ 0 = k l ν 0 + m n   , with ( k , l ) = 1   . If k = 0   , then μ 0 Q   , and we are done. Otherwise take a , b Z   such that a k + b l = 1   , so that ( l k a b ) GL 2 ( Z )   , and set ( μ 0 , ν 0 ) = ( l k a b ) ( μ 0 , ν 0 ) .   Then μ 0 = l ( k l ν 0 + m n ) + k ν 0 = l m n Q .   and ν 0 = a ( k l ν 0 + m n ) + b ν 0 = 1 l ν 0 + a m n Q ,   We now take ν = ν 0 / 2   and p / q = μ 0   , in lowest terms. By Remark  2.1  D c μ ν   and D p 2 q , ν c   are isomorphic.
Proposition 2.7. Let p   and q   be non-zero integers such that ( p , q ) = 1   , and let ν R   . Then D p 2 q , ν c   is Morita equivalent to D 0 , q ν c   .
  • Proof. By Remark  2.1 we may assume that p   and q   are positive. By applying Proposition  2.2 to ( μ , ν ) = ( q / 2 , ν )   , we get that D 0 , ν c = D q / 2 , ν c   is Morita equivalent to D 1 2 q , ν q c   , thus proving the proposition for p = 1   . For p > 1   , let r 0 = q   , r 1 = p   , and, if r i + 1 1   , define r i + 2   by r i = m i + 1 r i + 1 + r i + 2   , where 0 r i + 2 < r i + 1   , and m i + 1 Z   .
    Actually, r i + 2 > 0   : otherwise r i + 1   divides r i   , and it follows that r i + 1   divides r j   for all j i   . In particular, r i + 1   divides both p   and q   , which contradicts the fact that r i + 1 1   . Now, since r i + 1 < r i   , there is an index i 0   for which r i 0 = 1   .
    On the other hand, it follows from Proposition  2.2 that, for any real number κ   , D r i 2 r i 1 , κ c   is Morita equivalent to D r i 1 2 r i , κ r i 1 r i c   , which in turn is isomorphic to D r i + 1 2 r i , κ r i 1 r i c   .
    Thus D p 2 q , ν c = D r 1 2 r 0 , ν c   is Morita equivalent to D r j 2 r j 1 , q ν r j 1 c   for any j i 0   . In particular, for j = i 0   , we have that D p 2 q , ν c   is Morita equivalent to D 1 2 r i 0 1 , ν q r i 0 1 c   , which, as shown above, is Morita equivalent to D 0 , ν q c   .
Theorem 2.8. Two quantum Heisenberg manifolds D c μ ν   , D μ ν c   are Morita equivalent if and only if c = c   and there exists a positive real number r   such that Z + 2 μ Z + 2 ν Z = r ( Z + 2 μ Z + 2 ν Z ) .   In particular, the rank of the free abelian group G μ ν = Z + 2 μ Z + 2 ν Z   is the same for Morita equivalent quantum Heisenberg manifolds, and:
  • (1) If rank G μ ν = 1 =   rank G μ ν   , then D c μ ν   is Morita equivalent to D μ , ν c   . In particular, D c μ ν   is Morita equivalent to the commutative Heisenberg manifold D 0 , 0 c   .
  • (2) If rank G μ ν = 2 =   rank G μ ν   , let { α , p q }   and { α , p q }   be bases of G μ ν   and G μ ν   , respectively, where α   and α   are irrational numbers and p , p , q , q Z   , ( p , q ) = ( p q ) = 1   . Then D μ , ν c   and D μ , ν c   are Morita equivalent if and only if there exists ( a b c d ) GL 2 ( Z )   such that q α = a q α + b c q α + d .   In particular, D p 2 q , ν c   is Morita equivalent to D 0 , q ν c   .
  • (3) If rank G μ ν = 3 =   rank G μ ν   , then D c μ ν   and D c μ ν   are Morita equivalent if and only if there exists ( a b c d e f g h i ) GL 3 ( Z )   such that 2 μ = 2 a μ + 2 b ν + c 2 g μ + 2 h ν + i and 2 ν = 2 d μ + 2 e ν + f 2 g μ + 2 h ν + i .  
  • Proof. It was shown in [Ab1,3.4that K 0 ( D c μ ν ) = Z 3 Z c   , which implies that D c μ ν   and D μ ν c   are not Morita equivalent for c c   .
    Besides ([Ab2,Theorem3.16), all tracial states on D c μ ν   induce the same homomorphism on K 0 ( D c μ ν )   , whose range is the group G μ ν = 2 μ Z + 2 ν Z + Z   .
    Since ([Rf1,2.2) there is a bijection between finite traces of Morita equivalent unital C *   -algebras we must have G μ ν = r G μ ν   for some positive real number r   when D c μ ν   and D c μ ν   are Morita equivalent. An immediate consequence of this fact is that the rank of G μ ν   is invariant under Morita equivalence.
    If rank G μ ν = 1   , then, by [Ab2,Remark3.5, D c μ ν   is isomorphic to D 0 , 1 2 p c   for some non-zero integer p   , so D c μ ν   is isomorphic to D 1 2 p , 0 c   by Remark  2.1 .
    Now, by Proposition  2.7 , D 1 2 p , 0 c   is Morita equivalent D 0 , 0 c   .
    If rank G μ ν = 2 =   rank G μ ν   and G μ ν = r G μ ν   for some positive r   , let { α , p q }   and { α , p q }   be bases of G μ ν   and G μ ν   , respectively, where α , α   are irrational numbers, and p , p , q , q   are integers, with ( p , q ) = ( p , q ) = 1   .
    Since Z G μ ν ( G μ ν )   we have that p ( p ) = ± 1   and, by Remark  2.1 , we may assume p = p = 1   . Then we have that α Z + 1 / q Z = r ( α Z + 1 / q Z )   , which implies that α q Z + Z = ( r q / q ) ( α q Z + Z )   . A standard argument shows that q α = a q α + b c q α + d for some ( a b c d ) GL 2 ( Z ) .   Therefore D q α 2 , 0 c   and D q α 2 , 0 c   are Morita equivalent by Corollary  2.3 . On the other hand, by Remark  2.6 , D c μ ν   and D μ , ν c   are isomorphic, respectively, to D m 2 n , β c   and D m 2 n , β c   , for some irrational numbers β   and β   and integers m , m , n , n   such that ( m , n ) = ( m , n ) = 1   . Therefore { 2 β , 1 n }   and { 2 β , 1 n }   are bases of G μ ν   and G μ ν   , respectively, and it follows from the argument above that D n β , 0 c   and D n β , 0 c   are Morita equivalent. It only remains to notice now that, by Proposition  2.7 and Remark  2.1 , D n β , 0 c   and D n β , 0 c   are Morita equivalent to D c μ ν   and D μ ν c   , respectively.
    Finally, if rank G μ ν = 3 =   rank G μ ν   and G μ ν = r G μ ν   for some positive r   , then let A = ( a b c d e f g h i ) GL 3 ( Z )   be the transpose of the matrix that changes coordinates between the bases { 2 r μ , 2 r ν , r }   and { 2 μ , 2 ν , 1 }   of G μ ν   . Then 2 μ = 2 a μ + 2 b ν + c 2 g μ + 2 h ν + i and 2 ν = 2 d μ + 2 e ν + f 2 g μ + 2 h ν + i ,   which implies, by Proposition  2.4 , that D c μ ν   and D c μ ν   are Morita equivalent.
References

  1. Abadie, B. [1995], Generalized fixed-point algebras of certain actions on crossed-products, Pacific J. Math. 171, 1-21.
  2. Abadie, B. [2000], The range of traces on quantum Heisenberg manifolds, Trans. Amer. Math. Soc. 352, 5767-5780.
  3. Abadie, B.; Exel, R. [1997], Hilbert C *   -bimodules over commutative C *   -algebras and an isomorphism condition for quantum Heisenberg manifolds, Rev. Math. Phys. 9, 411-423.
  4. Abadie, B.; Eilers, S.; Exel, R. [1998], Morita equivalence for crossed products by Hilbert C *   -bimodules, Trans. Amer. Math. Soc. 350, 3043-3054.
  5. Elliott, G.; Gong, G. [1996], On the classification of C *   -algebras of real rank zero, II, Ann. of Math. 144, 497-610.
  6. Kurosh, A.G. [1960], The theory of groups, Vol. 2, Chelsea Publishing Company, Second edition.
  7. Packer, J. [1987], C*-algebras generated by projective representations of the discrete Heisenberg group, J. Operator Theory 18, 41-66.
  8. Packer, J. [1988], Strong Morita equivalence for Heisenberg C*-algebras and the positive cones of their K 0   -groups, Canad. J. Math. XL, 833-864.
  9. Pimsner, M.V. [1997], A class of C *   -algebras generalizing both Cuntz-Krieger algebras and crossed products by Z   , Fields Inst. Comm. 12, AMS, 189-212.
  10. Rieffel, M. [1981], C*-algebras associated with irrational rotations, Pacific J. Math. 93, 415-429.
  11. Rieffel, M. [1982], Applications of strong Morita equivalence to transformation C*-algebras, Proc. Symp. Pure Math. 38 (Part 1), 299-310.
  12. Rieffel, M. [1983], The cancellation theorem for projective modules over irrational rotation C*-algebras, Proc. London Math. Soc. 3 (No 47), 285-302.
  13. Rieffel, M. [1989], Deformation quantization of Heisenberg manifolds, Comm. Math. Phys. 122, 531-562.
  14. Rieffel, M. [1990], Proper actions of groups on C*-algebras, Mappings of operator algebras, Proc. Japan-US joint seminar, Birkhäuser, 141-182.

Centro de Matematicas. Facultad de Ciencias. Igua 4225, CP 11 400, Montevideo, Uruguay. E-mail address : abadie@cmat.edu.uy