EXAMPLES OF NON-COMPACT QUANTUM GROUP ACTIONS

PIOTR M. SOLTAN

Abstract. We present two examples of actions of non-regular locally compact quantum groups
on their homogeneous spaces. The homogeneous spaces are defined in a way specific to these
examples, but the definitions we use have the advantage of being expressed in purely C∗ -algebraic
language. We also discuss continuity of the obtained actions. Finally we describe in detail a
general construction of quantum homogeneous spaces obtained as quotients by compact quantum
subgroups.

1. Introduction
The main aim of this paper is to present two examples of quantum homogeneous spaces for
non-regular quantum groups. We will construct quotients of the quantum “az + b” and quantum
E(2) groups (from [27, 18, 17] and [21] respectively) by their classical subgroups. In both cases our
definition of the quantum homogeneous space will be deeply rooted in the particular form of the
C∗ -algebras related to these quantum groups. On the other hand, our definitions of quantum ho-
mogeneous spaces will be given in purely C∗ -algebraic language, without any use of von Neumann
algebras. Let us also remark that non-regular quantum groups do not fit into the elaborate and
powerful framework of [20]. We will prove that in both cases the action of the quantum group on
its homogeneous space is continuous in an appropriate sense. The second example (presented in
Section 4) can also be used as basis of a simple definition of a quantum homogeneous space of the
form G/K, where G is a quantum group and K is a compact quantum subgroup of G. We develop
this idea in the last section. The notion of a quantum subgroup we will use is more restrictive
than that of [20]. On the other hand, our construction works for general bisimplifiable Hopf C∗ -
algebras, not only for locally compact quantum groups. In particular we can use universal versions
of quantum groups ([8], [19, Section 5]).
We will precede our examples of quantum homogeneous spaces with a brief discussion of actions
of classical locally compact groups on C∗ -algebras. We shall provide a description of such actions
in language appropriate for generalization to quantum groups. Continuity of actions is discussed
and a definition of a continuous action of a quantum group is proposed and compared with existing
approaches.
The tools used to prove our results will range from the notion of G-product and Lanstad
algebra to C∗ -algebras generated by unbounded affiliated elements. The material on crossed
products by of C∗ -algebras by actions of abelian groups (including the notion of a G-product
and Lanstad algebra) can be found in [11, 13] and [7]. For the notions of multiplier algebra of a
C∗ -algebra, strict topology, morphisms of C∗ -algebras and elements affiliated to C∗ -algebras we
refer to [24, 10, 25]. Let us only mention here that by a morphism form a C∗ -algebra A to another
C∗ -algebra B we shall mean a ∗-homomorphism ϕ form A to the multiplier algebra M(B) of B
which is nondegenerate, i.e. one whose image contains an approximate unit for B. The set of all
morphisms from A to B will be denoted by Mor(A, B). In [25] the reader will also find a detailed
exposition of the concept of a C∗ -algebra generated by unbounded elements affiliated with it. We
will also use some earlier work on one of our our examples from [16] as well as a wide range of
results related to the considered quantum groups ([22, 21, 23, 27, 18, 17]).

Date: April 23, 2010.

2000 Mathematics Subject Classification. 46L89, 46L55, 46L85.
Key words and phrases. quantum group, quantum group action, C∗ -algebra, quantum homogeneous space.
Research partially supported by Polish government grant no. N201 1770 33, European Union grant PIRSES-
GA-2008-230836 and Polish government matching grant no. 1261/7.PR UE/2009/7.
1
2 PIOTR M. SOLTAN

The definition of a locally compact quantum group is due to Kustermans and Vaes ([9]), but
we will not be making any use of the Haar weights of the quantum groups under consideration.
On the other hand our quantum groups will have a continuous counit. For the theory of compact
quantum groups we refer to [26].
Let us briefly describe the contents of the paper. We begin with a section which carefully
rephrases the theory of actions of locally compact groups on C∗ -algebras int the language suitable
for quantum groups (much like various aspects of locally compact group actions on locally compact
spaces were treated in [5]). There are no new results in that section, but we feel that some concepts
present in the literature require clarification. At the end of this section we define continuous
actions of quantum groups on C∗ -algebras (quantum spaces) and provide a short explanation of
the standard problems encountered in construction of quantum homogeneous spaces. In Section
3 we introduce the quantum “az + b” groups for various values of the deformation parameter and
define the homogeneous space obtained as the quotient by a classical subgroup. This construction
was already performed in [16]. Then we analyze the action of the quantum “az + b” group on its
homogeneous space to show that it is continuous. The example from Section 4 is very similar in
spirit, but the methods of analyzing this example are significantly different.
Motivated by the Example from Section 4, in the last section we generalize known construction
of a quotient quantum homogeneous space for compact quantum group (cf. [14, 15]) to the situation
where the group is no longer compact, but the subgroup is. We show that the action of the original
quantum group on the resulting quantum homogeneous space is continuous.

2. Continuity of classical actions

Let (B, G, α) be a C∗ -dynamical system. Then for each x ∈ B we have the function
α(x) : G 3 t 7−→ αt (x) ∈ B. (2.1)
This function has constant norm and is norm-continuous by assumption. In particular it does not
vanish at infinity on G unless x = 0. It follows that α(x) 6∈ C0 (G) ⊗ B
for non-zero x. However,
it is easy to see that α is a ∗-homomorphism from B to M C0 (G) ⊗ B , where the last algebra
is naturally identified with the space of all norm-bounded functions G → M(B) continuous in the
strict topology.
This raises the question how to describe the special property that for each x ∈ B the function
α(x) has its values in B (not merely in M(B)) and is norm-continuous (not only strictly contin-
uous). The well known answer is that these properties are equivalent to the fact that for each
f ∈ C0 (G) the element
(f ⊗ 1)α(x) ∈ C0 (G) ⊗ B
(consider a function f constant and non-zero on a neighborhood of a given point t ∈ G).
Let us note here the fact which is well known, but difficult to find in the literature:
Proposition 2.1. Let (B, G, α) be a C∗ -dynamical system and for each x ∈ B let α(x) be defined
by (2.1). Then the linear span of the set
(f ⊗ 1)α(x) f ∈ C0 (G), x ∈ B


is dense in C0 (G) ⊗ B.
We will give a very short and easy proof of this fact once appropriate structure on C0 (G) has
been defined (in the proof of Proposition 2.3). Nevertheless we want to include an elementary
proof of Proposition 2.1 which does not use the additional structure of C0 (G).

Proof of Proposition 2.1. We need to approximate any element of C0 (G) ⊗ B by functions of a

specific form. It is enough to approximate elements from a dense subspace Cc (G, B) ⊂ C0 (G, B) ∼
=
C0 (G) ⊗ B. Therefore let F be a fixed continuous function G → B with compact support. For
t ∈ supp F let


Ut = s ∈ G F (s) − αst−1 F (t) < ε .
 EXAMPLES OF NON-COMPACT QUANTUM GROUP ACTIONS 3

n
S
By compactness of supp F there exist t1 , . . . , tn such that supp F ⊂ Uti . Let (χi )i=1,...,n be a
i=1
partition of unity subordinated to the covering (Uti )i=1,...,n of supp F and define
n 
X

X(t) = χj (t)αt αt−1 F (tj ) .
j
j=1

In other words
n 

(χj ⊗ 1)α αt−1 F (tj ) .
X
X=
j
j=1
Take now t ∈ supp F . We have
n
X

X(t) − F (t) = χj (t)αtt−1 F (tj ) − F (t)
j
j=1
Xn n

X
= χj (t)αtt−1 F (tj ) − χj (t)F (t)
j
j=1 j=1
n
X

≤ χj (t) αtt−1 F (tj ) − F (t) .
j
j=1

Now the j-th term of the last sum is non-zero only if t ∈ Utj . It follows that each term of this
sum is either zero or less than χj (t)ε.
We have thus shown that for each i we have sup X(t)−F (t) ≤ ε. On the other hand, outside
t∈Uti
n
S
Uti both X and F are zero. 
i=1

An immediate corollary of Proposition 2.1 is that the linear span of


(f ⊗ y)α(x) f ∈ C0 (G), x, y ∈ B
is dense in C0 (G) ⊗ B. Indeed, if f ⊗ y is a simple tensor in C0 (G) ⊗ B then we can approximate
it by finite sums of the form
(fi ⊗ 1)α(xi ).
X

Now if (eλ ) is a (bounded) approximateP unit for B then, by the diagonal argument, we can
approximate f ⊗ y by sums of the form (fi ⊗ eλ )α(xi ) because
X 
(fi ⊗ eλ )α(xi ) − g ⊗ y ≤ (1 ⊗ eλ ) (fi ⊗ 1)α(xi ) − g ⊗ y
X
+ g ⊗ (eλ y − y) .

Since the approximation works for simple tensors, it also works for general elements of C0 (G) ⊗ B.
Corollary 2.2. Let (B, G, α) be a C∗ -dynamical

system and for each x ∈ B let α(x) be defined
by (2.1). Then α ∈ Mor B, C0 (G) ⊗ B .
Let us note here one other density condition appearing in the literature. Quite clearly the linear
span of Z 
ϕ(t)αt (x) dt ϕ ∈ L1 (G), x ∈ B

is dense in B (any x ∈ B is the limit of such integrals with δ-like net of integrable functions). In
order to rewrite this condition in a more convenient way let us introduce for each ϕ ∈ L1 (G) the
continuous functional Z
ωϕ : C0 (G) 3 f 7−→ ϕ(t)f (t) dt.
Then we see that the linear span of
(ωϕ ⊗ id)α(x) ϕ ∈ L1 (G), x ∈ B


is dense in B.
4 PIOTR M. SOLTAN

Let us introduce the standard comultiplication ∆ ∈ Mor C0 (G), C0 (G) ⊗ C0 (G) :
∆(f )(s, t) = f (st). (2.2)
Then one easily checks that (∆ ⊗ id) ◦ α = (id ⊗ α) ◦ α.
Clearly one can also see that ( ⊗ id) ◦ α = id, where  is the evaluation functional
C0 (G) 3 f 7−→ f (e) ∈ C.
We will now state and prove a simple proposition dealing with various conditions defining
continuity of a group action in the language adaptable to the more general context of quantum
groups.
Proposition 2.3. Let G be a locally compact group and let B be a C∗ -algebra. Let


α ∈ Mor B, C0 (G) ⊗ B
be such that
• with ∆ defined by (2.2) we have
(∆ ⊗ id) ◦ α = (id ⊗ α) ◦ α (2.3)
• for any f ∈ C0 (G) and x ∈ B we have
(f ⊗ 1)α(x) ∈ C0 (G) ⊗ B,
Then the following conditions are equivalent
(1) ( ⊗ id) ◦ α = id, where  is the evaluation at the neutral element of G,
(2) ker α = {0},
(3) the linear span of
(f ⊗ 1)α(x) f ∈ C0 (G), x ∈ B

(2.4)
is dense in C0 (G) ⊗ B,
(4) the linear span of
(ωϕ ⊗ id)α(x) ϕ ∈ L1 (G), x ∈ B

(2.5)
is dense in B.
Moreover, if the equivalent conditions (1)–(4) are satisfied then then there exists a continuous
action α of G on B such that α is defined by (2.1).
Proof. Let us begin with defining a family (αt )t∈G of maps B → B by
αt (x) = α(x)(t) = (δt ⊗ id)α(x),
where δt is the evaluation functional C0 (G) 3 f 7→ f (t). One immediately find that each αt is
an endomorphism of B and that for each x ∈ B the mapping G 3 t 7→ αt (x) is norm continuous.
Moreover, it follows from (2.3) that
αt ◦ αs = αts
for all t, s ∈ G. In particular αe is an idempotent mapping B → B which commutes with all αt ’s.
The range of all αt ’s is equal to the range of αe and their kernels are all equal to ker αe .
It follows that (1)⇔(2). Indeed, (1) clearly implies (2), but (2) means that for any non-zero
x ∈ B the function t 7→ αt (x) is non zero. Thus if x ∈ ker αe then αt (x) = 0 for all t and so x = 0.
In other words, αe is an idempotent with zero kernel, so αe = id. This is exactly (1).
In particular if (1) is satisfied then all αt ’s are automorphisms and α becomes a continuous
action of G on B. To see that this implies (3) without using Proposition 2.1 note that the mapping
C0 (G) ⊗alg B 3 (f ⊗ x) 7−→ (f ⊗ 1)α(x) ∈ C0 (G) ⊗ B (2.6)
extends to an automorphism of C0 (G) ⊗ B. Indeed, the map (2.6) is bounded and multiplicative
(because C0 (G) is commutative). The inverse mapping is given by
(f ⊗ x) 7−→ (f ⊗ 1)(id ⊗ κ) α(x) ,

where κ is the automorphism of C0 (G) given by κ(f )(t) = f (t−1 ). Therefore the linear span of
(2.4) is dense in C0 (G) ⊗ B as the image of C0 (G) ⊗alg B under an automorphism of C0 (G) ⊗ B.
 EXAMPLES OF NON-COMPACT QUANTUM GROUP ACTIONS 5

The fact that (3)⇒(4) is standard: approximate a simple tensor f ⊗ y ∈ C0 (G) ⊗ B by sums of
the form
(fi ⊗ 1)α(xi ).
X

Then take ϕ ∈ L1 (G) such that ωϕ (f ) = 1. Then the elements

X
(ωϕfi ⊗ id)α(xi )
approximate y and consequently the linear span of (2.5) is dense in B.
Finally let us see that (4) implies (1). If (1) is not satisfied then we know that the ranges of
all αt ’s are equal to the range of αe which is strictly contained in B and closed (as the image of a
C∗ -algebra under a ∗-homomorphism). Then it follows that for any ϕ ∈ L1 (G) the element
Z
(ωϕ ⊗ id)α(x) = ϕ(t)αt (x) dt

also lies in the range of αe and we see that (4) is not satisfied. 
In what follows we will deal with quantum groups of the form G = (A, ∆) with continuous counit
. We will say that G acts continuously on a C∗ -algebra B if there is a morphism α ∈ Mor(B, A⊗B)
such that
• (∆ ⊗ id) ◦ α = (id ⊗ α) ◦ α,
• for any c ∈ A and x ∈ B we have (c ⊗ 1)α(x) ∈ A ⊗ B,
• ( ⊗ id) ◦ α = id.
In the literature the last condition is often replaced by either requirement that the linear span
of
(c ⊗ 1)α(x) c ∈ A, x ∈ B


be dense in A ⊗ B or that 
(ω ⊗ id)α(x) ω ∈ A∗ , x ∈ B
be dense in B (where A∗ is the space of normal linear functionals on A, cf.[19]). In [3, Proposition
5.8] it is shown that these conditions are equivalent for regular locally compact quantum groups
and that the second condition is strictly weaker for semi-regular, but non-regular locally compact
quantum groups.
If α ∈ Mor(B, A ⊗ B) is such that (∆ ⊗ id) ◦ α = (id ⊗ α) ◦ α and ker α = {0} then the
established terminology is that α is a reduced action of G on B (see e.g. [20, 6]). As Proposition
2.3 shows, in case when G is a classical locally compact group, all these conditions are equivalent.
Let G = (A, ∆) and K = (C, ∆C ) be quantum groups. We say that a morphism π ∈ Mor(A, C)
identifies K as a closed quantum subgroup of G if π is a surjective ∗-homomorphism A → C and
(π ⊗ π) ◦ ∆ = ∆C ◦ π. As we mentioned in Section 1, such a definition of a quantum subgroup
is more restrictive than that of [20]. In the following sections we will give examples of situations
where this notion of quantum subgroup applies.
Let us explain the difficulty which we come across as we try to generalize the standard construc-
tion of a quotient quantum homogeneous space from [15]. Let us assume that the C∗ -algebras
A and B are commutative, so that A = C0 (G) and C = C0 (K), where G is a locally com-
pact group and K is a closed subgroup of G. In this situation π is the restriction morphism
A 3 f 7→ f K ∈ M(C). The range rarely coincides with C. In the language of noncommutative
topology the homogeneous space G/K is described by the C∗ -algebra B = C0 (G/K). It is a
nontrivial question how to identify this C∗ -algebra as a subalgebra of M(A) = Cb (G). Clearly any
element x ∈ B should satisfy
(id ⊗ π)∆(x) = x ⊗ 1, (2.7)
but this is certainly not enough. In fact all elements of M(B) also satisfy this condition. In the
examples of Sections 3 and 4 G = (A, ∆) will be a non-regular locally compact quantum group
with a subgroup K = (C, ∆C ) (which in both cases will be a classical group) and we will produce
a candidate for B. It will be a C∗ -subalgebra of M(B) consisting of elements x satisfying (2.7)
and some additional conditions tailored to these specific examples. The next question which arises
naturally after the definition of a homogeneous space is whether the restriction of ∆ to B actually
6 PIOTR M. SOLTAN

defines an action of G on G/K. We will have to check that α = ∆ B is a morphism from B to

A ⊗ B and that algebraic conditions of a (co)action are satisfied. More importantly we would like
to show that the action α is continuous. None of these facts is totally obvious and we prove them
by application of the theory of C∗ -algebras generated by unbounded elements ([25]).

3. The action of the quantum “az + b” group on its homogeneous space

The construction of the quantum “az + b” group starts with choosing the value of the deforma-
tion parameter q. There are at least three possibilities to choose q. They are described in detail
in [17]. One of these possibilities is to take q ∈]0, 1[ (cf. [27, Appendix A]). It is not important for
our purposes to dwell on this choice since the aspect of quantum “az + b” groups we shall need
are the same regardless of the value of the deformation parameter. All details can be found in
[27, 18, 17].
We let Γ be the subgroup of the multiplicative group C \ {0} generated by q and {q it t ∈ R}
with an appropriate fixed choice of logarithm of q. The group Γ is then self-dual and we denote
by χ the nondegenerate bicharacter Γ × Γ → T such that
χ(γ, γ 0 ) = χ(γ 0 , γ),
χ(γ, q it ) = |γ|it ,
χ(γ, q) = Phase γ
for all γ, γ 0 ∈ Γ and t ∈ R.
Then we let Γ be the closure of Γ in C and let β be the action of Γ on Γ by multiplication. We
define A = C0 (Γ) oβ Γ. An affiliated element b η A is introduced as the image of the canonical
generator of C0 (Γ) in A, while a η A
as the unique affiliated element such that Sp a ⊂ Γ, a is
invertible (with a−1 η A) and χ(a, γ) γ∈Γ is the canonical family of unitary elements of M(A)
implementing the action β on C0 (Γ).
By the results of [27, 18] (the actual reference depends on the value of q, cf. [17]) there exists
a unique ∆ ∈ Mor(A, A ⊗ A) such that
∆(a) = a ⊗ a, ∆(b) = a ⊗ b +̇ b ⊗ 1
and G = (A, ∆) is a quantum group (in fact a non-regular locally compact quantum group cf. also
[28]).
Let us first see that Γ is a subgroup
of G in the sense described at the end of Section 2, i.e. there
exists a morphism π ∈ Mor A, C0 (Γ) such that
∆ ◦ π = (π ⊗ π) ◦ ∆
and π is a surjection onto C0 (Γ). Indeed, since the C∗ -algebra A is generated by unbounded
elements a, a−1 and b affiliated with it (cf. [27, 18, 17]), it is enough to define π on the generators
by putting π(b) = 0 and letting π(a) be the canonical generator of C0 (Γ) (the identity function
Γ 3 γ 7→ γ ∈ C).
Let us remark here that in [16] it is the dual group of Γ which is identified with a subgroup of
G which we are describing. This is correct since Γ ∼ = Γ,
b but the distinction has no bearing on our
construction of quantum homogeneous space, so we will stick with the choice of Γ as a subgroup
of G.
We shall now define a quantum homogeneous space G/Γ. By definition the C∗ -algebra B of
continuous functions vanishing at infinity on G/Γ is the set of those x ∈ M(A) for which
(1) (id ⊗ π)∆(x) = x ⊗ 1,
(2) for any y ∈ C∗ (Γ) ⊂ M(A) the element yx ∈ A,
(3) the mapping Γ 3 γ 7→ χ(a, γ)∗ xχ(a, γ) is continuous.
Condition (1) selects elements of M(A) which are “constant along the fibers” of the fibration
G → G/Γ. However, they cannot correspond to functions vanishing at infinity on G/Γ (e.g. the
unit 1 satisfies this condition). Therefore condition (2) is introduced to select those “continuous
bounded functions on G” which not only are constant on cosets of Γ, but in addition vanish
at infinity “in the direction transversal to the cosets.” Condition (3) does not have a classical
 EXAMPLES OF NON-COMPACT QUANTUM GROUP ACTIONS 7

analogue. This definition of the algebra of “continuous functions vanishing at infinity on G/Γ”
was first introduced in [16]. 
It is not difficult to check that B = f (b) f ∈ C0 (Γ) . This is because conditions (1)–(3)
are really the Lanstad conditions for the Γ-product structure on A = C0 (Γ) oβ Γ (cf. [11, 13]).
Therefore the homogeneous space G/Γ is a classical space, i.e. one described by a commutative
C∗ -algebra.
It was pointed out already in [16] that the restriction of ∆ to B ⊂ M(A) defines a morphism
α ∈ Mor(B, A ⊗ B) and that
(∆ ⊗ id) ◦ α = (id ⊗ α) ◦ α.
Moreover one can easily see that ( ⊗ id) ◦ α = id, where  is the counit of G defined by (a) = 1,
(b) = 0. Therefore, in order to say that the action of G on the homogeneous space G/Γ is
continuous we only need to prove the following:
Theorem 3.1. Let G = (A, ∆) be the quantum “az +b” group and let B ⊂ M(A) be the C∗ -algebra
of functions vanishing at infinity on the quantum homogeneous space G/Γ. Let α ∈ Mor(B, A⊗B)
be the morphism defined above. Then for any c ∈ A and any x ∈ B the element
(c ⊗ 1)α(x) ∈ A ⊗ B. (3.1)
Proof. We shall identify B with C0 (Γ) and A ⊗ B with A ⊗ C0 (Γ) ∼
= C0 (Γ, A). The element b η B
corresponds to the function Γ 3 z 7→ z ∈ C and α(b) to
Γ 3 z 7−→ (az +̇ b) η A.
(cf. [25, Formula 2.6]).
It follows that for any f ∈ C0 (Γ) the image of f (b) under α is identified with the element
Γ 3 z 7−→ f (az +̇ b)
strict


of Cb Γ, M (A) . Therefore with x = f (b) the element (3.1) corresponds to the function
Γ 3 z 7−→ cf (az +̇ b)
which is an element of Cb (Γ, A). We want to show that this is in fact an element of C0 (Γ, A) =
C0 (Γ) ⊗ A.
According to results [22, 27, 18] (again the reference depends on the choice of q) there is a
continuous function Fq : Γ → T such that az +̇ b = Fq (z −1 ba−1 )∗ azFq (z −1 ba−1 ), so that
f (az +̇ b) = Fq (z −1 ba−1 )∗ f (az)Fq (z −1 ba−1 ). (3.2)
Our aim is to show that for any fixed c ∈ A the norm of
cFq (z −1 ba−1 )∗ f (az)Fq (z −1 ba−1 )
tends to 0 as z goes to ∞ of Γ. Clearly this norm is equal to the norm of
cFq (z −1 ba−1 )∗ f (az).
Since ba−1 η A, it is known (see e.g. [27, Theorem 5.1]) that Fq (z −1 ba−1 ) tends to 1 strictly in
M(A) az z → ∞ (note that Fq (0) = 1). Therefore for any ε > 0 there exists a constant Rε such
that
cFq (z −1 ba−1 )∗ − c < ε.
for all z ∈ Γ with |z| > Rε . Thus the norm
cFq (z −1 ba−1 )∗ f (az) ≤ cf (az) + cFq (z −1 ba−1 )∗ − c f (az) < cf (az) + εkf k.

To see that the norm of cf (az) goes to zero as z → ∞ we can assume that f has compact support
and that c is of the form
c = f˜(b)g(a)
for some f˜ ∈ C0 (Γ) and g ∈ C0 (Γ) (the set of such elements is linearly dense in A). Then
cf (za) = f˜(b)Gz (a),
8 PIOTR M. SOLTAN

where Gz (z 0 ) = g(z 0 )f (zz 0 ). Since g vanishes at 0, it is easy to see that kGz k −−−→ 0 and it
z→∞
follows that cf (za) < ε for sufficiently large |z|. 

4. The action of the quantum E(2) group on its homogeneous space

For the description of the quantum E(2) group we will fix a Hilbert space H with an orthonormal
basis (ei,j )i,j∈Z . We will change our notation slightly in order to be in agreement with conventions
established in the literature [21, 23]. Let q ∈]0, 1[ be a parameter and let Cq be the subgroup of
C \ {0} generated by q and {q it t ∈ R} — this is Γ from Section 3 for the case of real q. As before
we let Cq be the closure of Cq in C.
We will now define two operators on H. We let v be the unitary operator defined uniquely by
vei,j = ei−1,j for a i, j ∈ Z and let n be the closed linear operator on H such that the linear span
of the orthonormal basis (ei,j )i,j∈Z is a core of n and nei,j = q i ei,j+1 for all i, j ∈ Z. Then n is a
normal operator with Sp n ⊂ Cq . We define A as the closure of the set of finite linear combinations
X
fl (n)v l , (4.1)

where fl ∈ C0 (Cq ) for all l ∈ Z. With this definition A is a non-degenerate C∗ -subalgebra of

B(H) isomorphic to C0 (Cq ) oβ Z, where β is the natural action by multiplication by q (cf. [21]).
Moreover v ∈ M(A) and n η A and A is generated by n and v ([25, Example 4]).
It is shown in [21] that there exists a unique ∆ ∈ Mor(A, A ⊗ A) such that
∆(v) = v ⊗ v, ∆(n) = v ⊗ n +̇ n ⊗ v ∗
and G = (A, ∆) is a non-regular locally compact quantum group ([1]) called the quantum E(2)
group.

group T is a subgroup of G, i.e. we have a quantum group morphism π ∈
The classical
Mor A, C(T) defined uniquely by putting π(n) = 0 and letting π(v) be the canonical unitary
generator of C(T). We wish to describe the quantum homogeneous space G/T.
To that end let us define structure of a Z-product on A different from the standard one (arising
from the fact that A is the crossed product of C0 (Cq ) by Z). We define a representation
V : Z 3 n 7−→ Vn = v n ∈ M(A)
and action of T on A by automorphisms (β̃µ )µ∈T such that
β̃µ (v) = µv, β̃µ (n) = µ−1 n
for all µ ∈ T. The triple (A, V, β̃) is a Z-product. This Z-product structure on A is not a “twist”
of the standard Z-product structure on A as defined in [7], but let us note that if we denote by β̂
the action of T on A dual to β then for each µ ∈ T.
β̃µ (y) = Uµ β̂µ (y)Uµ∗ ,
where Uµ is the unitary operator such that Uµ ei,j = µ−j ei,j .
Following the example from Section 3 we define the C∗ -algebra of B of continuous functions
vanishing at infinity on G/T as the set of those x ∈ M(A)
(1) (id ⊗ π)∆(x) = x ⊗ id,
(2) for any y ∈ C∗ (Z) ⊂ M(A) the element yx ∈ A.
The above conditions are, as in Section 3, precisely the Lanstad conditions for the Z-product
structure we have defined (in particular B really is a C∗ -algebra). Condition 1 is exactly the
condition of being β̃-invariant, while condition 2 says that the “function” x vanishes at infinity
in the direction transversal to that of the subgroup T. In fact condition (2) is equivalent to
demanding that x ∈ A (because C∗ (Z) is unital). The third Lanstad condition is empty because
Z is a discrete group.
Before proceeding let us define a certain C∗ -algebra which will turn out to be the algebra
of continuous unctions vanishing at infinity on G/T. Let C0 (Z ∪ {+∞}) be the C∗ -algebra of
sequences (xn )n∈Z for which lim xn = 0 and lim xn exists and is finite. We let C be the
n→−∞ n→+∞
 EXAMPLES OF NON-COMPACT QUANTUM GROUP ACTIONS 9

crossed product C = C0 (Z ∪ {+∞}) o Z, where the action of Z is by translation. The C∗ -algebra

C is an extension of K by C(T). More precisely we have
  
x y
C∼
= x ∈ T , y, z, u ∈ K
z u
(where T is the Toeplitz algebra) and the extension is

0 /K∼= M2 (K) /C ρ
/ C(T) /0
x y
with ρ sending a matrix z u to the symbol of x.
Proposition 4.1. Let B be the algebra of continuous functions vanishing at infinity on G/T.
Then
(1) The C∗ -algebra B is the closure of the set of finite sums of the form
X
fk (n)v k (4.2)

where for each k ∈ Z the function fk ∈ C0 (Γ) is such that

fk (µz) = µk fk (z)
for all z ∈ Γ and µ ∈ T.
(2) The operator vn is affiliated with B,
(3) B is generated by vn.
(4) B is isomorphic to C.
Proof. We shall first prove point (1). Let B0 be the closure of the set of finite sums described in
Statement (1). Since each term fk (n)v n is invariant under β̃ and is already contained in A (as
opposed to M(A)) we have B0 ⊂ B. Moreover B0 is invariant under conjugation with v. Finally
let us note that
C∗ (Z)B0 C∗ (Z) (4.3)
is dense in A (as before, we identify C (Z) with the C -subalgebra of M(A) generated by v). This
∗ ∗

is because the functions fk appearing in (4.2) are trigonometric monomials along the individual
circles of Cq . Therefore we can uniformly approximate on each circle sufficiently regular functions
(e.g. twice differentiable) and those are dense in all continuous functions on the circle. This can
be done on each circle separately. As a result all elements of the form (4.1) lie in the closure of
(4.3). By [7, Lemma 2.6] B0 = B.
Now we shall deal with points (2)–(4). The operator vn is affiliated with B. This is because
the z-transform of vn clearly belongs to B0 and the element 1 − zvn ∗
zvn = (1 + n∗ n)−1 is a strictly
positive element of B (cf. [25, Section 1]). Let us denote by X the operator vn and let X = U |X|
be the polar decomposition of X. We have U ei,j = ei−1,j+1 and |X|ei,j = q i ei,j . One easily finds
that for fk ∈ C0 (Cq ) such that fk (µz) = µk fk (z) we have
fk (n)v k = fk |X| U k .


(4.4)
A moment of reflection shows now that B ∼ = C and point (4) follows.
Let ρ be a representation of B on a Hilbert space L. Then ρ(vn) is a closed operator on
L. Moreover by (4.4) the image under ρ of any element x ∈ B0 is determined by the polar
decomposition of ρ(vn). This means that vn separates representations of B and it follows that B

−1
is generated by vn by [25, Theorem 3.3], since 1 + (vn)∗ (vn) ∈ B. 
Before proceeding let us emphasize the fact that B ⊂ A. Moreover since B ⊂ A we have that
B = x ∈ A β̃µ (x) = x for all µ ∈ T


= x ∈ A (id ⊗ π)∆(x) = x ⊗ 1 .


We also have
Lemma 4.2. A ⊗ B = X ∈ A ⊗ A (id ⊗ β̃µ )(X) = X for all µ ∈ T .

10 PIOTR M. SOLTAN

Proof. The inclusion “⊂” is obvious. For the converse one let us take X ∈ A ⊗ A such that
(id ⊗ β̃µ )(X) = X for all µ ∈ T. X is a limit of finite sums of simple tensors:
Nk
(k) (k)
X
X = lim ai ⊗ bi .
k→∞
i=1
Let dµ be the normalized Haar measure on T. We have
Z Nk Z
(k) (k)
X
X = (id ⊗ β̃µ )(X) dµ = lim ai ⊗ β̃µ (bi ) dµ
k→∞
i=1
(k) R (k)
Clearly the elements ci = β̃µ (bi ) dµ are β̃-invariant, so that they belong to B. It follows that
X ∈ A ⊗ B. 
Now let us describe the action of G on its homogeneous space G/T. The image of vn η A under
∆ is
v 2 ⊗ vn +̇ vn ⊗ 1.
By [29, Theorem 6.1] and [22, Theorem 5.1] this element is affiliated with A⊗B. Since vn generates
B, we see that the map α = ∆ B is a morphism form B to A ⊗ B. Clearly we have
(∆ ⊗ id) ◦ α = (id ⊗ α) ◦ α
and ( ⊗ id) ◦ α = id, where  is the counit of G (defined by (v) = 1 and (n) = 0).
Since B ⊂ A we have α(B) ⊂ A ⊗ A by [23, Formula (57)]. Note also that
(id ⊗ β̃µ ) ◦ ∆ = ∆ ◦ β̃µ . (4.5)
In view of Lemma 4.2, this implies that in fact α(x) ∈ A ⊗ B for any x ∈ B. In particular for any
c ∈ A and x ∈ B we have (c ⊗ 1)α(x) ∈ A ⊗ B. Thus we obtain
Corollary 4.3. α ∈ Mor(B, A ⊗ B) is a continuous action of G on G/T.
Remark 4.4. Let us also note that the set of all elements (c ⊗ 1)α(x) with c ∈ A and x ∈ B is not
only contained (as shown above),Rbut also linearly dense in A ⊗ B. Indeed if we denote by E the
conditional expectation A 3 c 7→ β̃µ dµ ∈ B used in the proof of Lemma 4.2 then it follows from
(4.5) that (id ⊗ E) ◦ ∆ = ∆ ◦ E. Using the fact that the span of (a ⊗ 1)∆(b) a, b ∈ A is dense
in A ⊗ A we see that
span (c ⊗ 1)α(x) c ∈ A, x ∈ B


= span (c ⊗ 1)∆(x) c ∈ A, x ∈ B


= span (c ⊗ 1)∆ E(d) c, d ∈ A



= span (c ⊗ 1)(id ⊗ E) ∆(d) c, d ∈ A



= span (id ⊗ E) (c ⊗ 1)∆(d) c, d ∈ A



is dense in A ⊗ B (cf. Lemma 4.2).

5. Quotients by compact subgroups

A pair G = (A, ∆) consisting of a C∗ -algebra A and a coassociative ∆ ∈ Mor(A, A ⊗ A) such
that
span ∆(a)(1 ⊗ a0 ) a, a0 ∈ A = A ⊗ A,

(5.1a)
span (a ⊗ 1)∆(a0 ) a, a0 ∈ A = A ⊗ A.

(5.1b)
is usually called a bisimplifiable Hopf C∗ -algebra (see e.g. [2]), while in [12] the such objects are
called “proper C∗ -bialgebras with cancellation property.” We will stick to the former terminology.
This section is devoted to the proof of the next theorem which is a direct generalization of
the construction in [14][Section 6], [15, Section 1] (cf. also [4]) to the situation where the original
quantum group (or Hopf C∗ -algebra) G is not compact. We include the assumption that G possesses
 EXAMPLES OF NON-COMPACT QUANTUM GROUP ACTIONS 11

a continuous counit in order to fit the scheme of earlier sections. In case G does not have a counit
the all statements of Theorem 5.1 (except (3)) are still true.
Note also that many of the results of Section 4 are in fact consequences of Theorem 5.1, but
the proofs given in Section 4 are different and independent of the next result.
Theorem 5.1. Let G = (A, ∆) be a bisimplifiable Hopf C∗ -algebra with a continuous counit 
and let K = (C, ∆C ) be a compact quantum group. Let π be a ∗-homomorphism A → C which is
surjective and satisfies
(π ⊗ π) ◦ ∆ = ∆C ◦ π.
Let B = x ∈ A (id ⊗ π)∆(x) = x ⊗ 1 . Then


(1) B is a nondegenerate C∗ -subalgebra of A,

(2) the map α = ∆|B has values in M(A ⊗ B) and belongs to Mor(A, A ⊗ B),
(3) ( ⊗ id) ◦ α = id,
(4) for any b ∈ B and a ∈ A we have (a ⊗ 1)α(b) ∈ A ⊗ B.
Proof. Denote γ = (id ⊗ π)∆ ∈ Mor(A, A ⊗ C). We note the identities
(γ ⊗ id) ◦ γ = (id ⊗ ∆C ) ◦ γ, (5.2a)
(∆ ⊗ id) ◦ γ = (id ⊗ γ) ◦ ∆ (5.2b)
which follow directly form the coassociativity of ∆. Next we note that
span γ(a)(1 ⊗ c) a ∈ A, c ∈ C = A ⊗ C,

(5.3a)
span (a ⊗ 1)γ(a0 ) a, a0 ∈ A = A ⊗ C

(5.3b)
(in particular both sets on the left hand side are contained in A ⊗ C). Both (5.3a) and (5.3b)
follow from (5.1) by the surjectivity of π. If K has a continuous counit then it can be shown that
 = C ◦ π and then (id ⊗ C ) ◦ γ = id. This means that γ defines a continuous (right) action of K
on A.
Let h be the Haar measure of K and let
E : A 3 a 7−→ (id ⊗ h)γ(a).
Here we use the strictly continuous extension of (id ⊗ h) to M(A ⊗ C). By (5.3a) E has values in
A. Indeed we have that h = vh0 for some h0 ∈ C ∗ and v ∈ C, so that
(id ⊗ h)γ(a) = (id ⊗ h0 ) γ(a)(1 ⊗ v)

is an element of A. Moreover E 2 = E because

E 2 (a) = (id ⊗ h ⊗ h) (γ ⊗ id)γ(a)




= (id ⊗ h ⊗ h) (id ⊗ ∆C )γ(a)
= (id ⊗ h)γ(a) = E(a).
In the above calculation we must remember that we are dealing with extensions of strictly con-
tinuous maps to multiplier algebras. For example the identity (h ⊗ h) ◦ ∆C = h follows from the
definition of the Haar measure, but (id ⊗ h ⊗ h) ◦ (id ⊗ ∆C ) = (id ⊗ h) is an identity between maps
on M(A ⊗ C). We obtain it by extending both sides from A ⊗ C to M(A ⊗ C) by strict continuity.
Similarly we show that E(A) ⊂ B. Indeed, denoting by H the map C 3 c 7→ h(c)1 we have



γ E(a) = γ (id ⊗ h)γ(a)


= (id ⊗ id ⊗ h) (γ ⊗ id)γ(a)


= (id ⊗ id ⊗ h) (id ⊗ ∆C )γ(a) (5.4)


= id ⊗ [(id ⊗ h) ◦ ∆C ] γ(a)
= (id ⊗ H)γ(a) = E(a) ⊗ 1.
As before the identities satisfied by the maps involved in the above calculation are extended by
strict continuity to multiplier algebras.
12 PIOTR M. SOLTAN

An immediate consequence of (5.4) is that B = E(A). To prove point (1) of Theorem 5.1 we
note that it follows from (5.3b) that
span ab a ∈ A, b ∈ B = span aE(a0 ) a, a0 ∈ A
 

= span a(id ⊗ h)γ(a0 ) a, a0 ∈ A



= span (id ⊗ h) (a ⊗ 1)γ(a0 ) a, a0 ∈ A



is a dense subset of A. In other words the inclusion of B into A belongs to Mor(B, A).
In particular M(B) is the idealizer of B in M(A) and M(A ⊗ B) is the idealizer of A ⊗ B in
M(A ⊗ B) ([10, Proposition 2.3]), i.e.

M(A ⊗ B) = X ∈ M(A ⊗ A) XY , Y X ∈ A ⊗ B for all Y ∈ A ⊗ B . (5.5)
Let us also note that we have
A ⊗ B = Y ∈ A ⊗ A (id ⊗ γ)(Y ) = Y ⊗ 1 .

(5.6)
Indeed, the inclusion “⊂” is clear. For “⊃” we use the conditional expectation E in the same
way as in the proof of Lemma 4.2. Any Y belonging to the right hand side of (5.6) satisfies
(id ⊗ E)(Y ) = Y and it is not difficult to see that A ⊗ B is the image of (id ⊗ E). Using (5.6) we
can show that
M(A ⊗ B) = X ∈ M(A ⊗ A) (id ⊗ γ)(X) = X ⊗ 1 .

(5.7)
The inclusion “⊃” follows because if X ∈ M(A ⊗ A) satisfies (id ⊗ γ)(X) = X ⊗ 1 then for any
Y ∈ A ⊗ B we have XY, Y X ∈ A ⊗ B, so that X ∈ M(A ⊗ B). Conversely, if X ∈ M(A ⊗ B) then
for any Y ∈ A ⊗ B we have XY ∈ A ⊗ B, so by (5.6) we have
(id ⊗ γ)(X) (Y ⊗ 1) = (id ⊗ γ)(X) (id ⊗ γ)(Y )

= (id ⊗ γ)(XY ) = (XY ⊗ 1) = (X ⊗ 1)(Y ⊗ 1).

This means that T = (id ⊗ γ)(X) − X ⊗ 1 ∈ M(A ⊗ A ⊗ B) satisfies
T (Y ⊗ 1) = 0
for all Y ∈ A ⊗ B. It follows that T (a ⊗ a0 ⊗ b) is zero for all a, a0 ∈ A, b ∈ B, so that T = 0 and
we obtain “⊂” in (5.7).
Let us now address point (2) of Theorem 5.1. The map α = ∆ B is a composition of the
inclusion of B into A with ∆. Therefore it is an element of Mor(B, A ⊗ A). We want to show
that it belongs to Mor(B, A ⊗ B). For this it is enough to demonstrate that the range of α lies in
M(A ⊗ B).
By application of (id ⊗ id ⊗ h) to both sides of (5.2b) we obtain
∆ ◦ E = (id ⊗ E) ◦ ∆.
This shows that the image of α is in the image of (id ⊗ E) extended to M(A ⊗ A). Clearly for any
X ∈ A ⊗ A and Y ∈ A ⊗ B we have
(id ⊗ γ) (id ⊗ E)(X) Y = XY ⊗ 1.


(5.8)
Both sides of this formula are strictly continuous with respect to X, so (5.8) holds also for X ∈
M(A ⊗
A). In view of (5.7), formula (5.8) (together with the analogous one (id ⊗ γ) Y (id ⊗
E)(X) = Y X ⊗ 1 and its extension) for proves that the range of α is contained in M(A ⊗ B).
Point (3) of Theorem 5.1 is a straightforward consequence of the definition of α. To prove point
(4) we note that
(a ⊗ 1)α(b) a ∈ A, b ∈ B = (a ⊗ 1)∆ E(a0 ) a, a0 ∈ A
 

= (a ⊗ 1)(id ⊗ E)∆(a0 ) a, a0 ∈ A


= (id ⊗ E) (a ⊗ 1)∆(a0 ) a, a0 ∈ A


so that (a ⊗ 1)α(b) ∈ A ⊗ B for all a ∈ A, b ∈ B. 

 EXAMPLES OF NON-COMPACT QUANTUM GROUP ACTIONS 13

Remark 5.2. Note that it follows easily from the last lines of the proof of Theorem 5.1 that the
linear span of the set
(a ⊗ 1)α(b) a ∈ A, b ∈ B


is dense in A ⊗ B (cf. Remark 4.4).

References
[1] S. Baaj: Répresentation régulière du groupe quantique Eµ (2) de Woronowicz. C. R. Acad. Paris, t. 314, Série
I, (1992), 1021–1026.
[2] S. Baaj & G. Skandalis: Unitaries muliplicatifs et dualité pour les poiduits croisés de C∗ -algébres. Ann. Sci-
ent. Éc. Norm. Sup. 4è série, t. 26 (1993), 425–488.
[3] S. Baaj, G. Skandalis & S. Vaes: Non-semi-regular quantum groups coming from number theory. Com-
mun. Math. Phys. 235 no. 1 (2003), 139–167.
[4] F. Boca: Ergodic actions of compact matrix pseudogroups on C∗ -algebras. In Recent Advances in Operator
algebras. Astérisque 232 (1995), pp. 93–109.
[5] D.A. Ellwood: A new characterization of principal actions. J. Funct. Anal. 173 (2000), 49–60.
[6] R. Fischer: Volle verschränkte Produkte für Quantengruppen und äquivariante KK-Theorie. Ph.D. thesis,
University of Münster, 2003.
[7] P. Kasprzak: Rieffel deformation via crossed products. J. Funct. Anal. 257 no. 5 (2009), 1288–1332.
[8] J. Kustermans: Locally compact quantum groups in the universal setting. Int. J. Math. 12 (2001), 289–338.
[9] J. Kustermans & S. Vaes: Locally compact quantum groups. Ann. Scient. Éc. Norm. Sup. 4è série, t. 33
(2000), 837–934.
[10] E.C. Lance: Hilbert C∗ -modules, a toolkit for operator algebraists. London Mathematical Society Lecture
Note Series 210, Cambridge University press 1995.
[11] M.B. Landstad: Duality theory for covariant systems. Trans. Amer. Math. Soc. 248, no. 2 (1979), 223–267.
[12] T. Masuda, Y. Nakagami & S.L. Woronowicz: A C∗ -algebraic framework for quantum groups. Int. J. Math.
14 no. 9 (2003) 903–1001.
[13] G.K. Pedersen: C∗ -algebras and their automorphism groups. Academic Press 1979.
[14] P. Podleś: Quantum spheres. Lett. Math. Phys. 14 (1987), 193–202.
[15] P. Podleś: Symmetries of quantum spaces. Subgroups and quotient spaces of quantum SU (2) and SO(3)
groups. Commun. Math. Phys. 170 (1995), 1–20.
[16] W. Pusz & P.M. Soltan: Analysis on a homogeneous space of a quantum group. Preprint, University of
Warsaw. arxiv:math.OA/0509610v3.
[17] W. Pusz & P.M. Soltan: On some low dimensional quantum groups. Preprint, University of Warsaw, to
appear in Quantum symmetries in noncommutative geometry P.M. Hajac editor. arxiv:0712.3369v1.
[18] P.M. Soltan: New quantum “az + b” groups. Rev. Math. Phys. 17 no. 3 (2005), 313–364.
[19] P.M. Soltan & S.L. Woronowicz: From multiplicative unitaries to quantum groups II. J. Funct. Anal. 252
no. 1 (2007), 42–67.
[20] S. Vaes: A new approach to induction and imprimitivity results. J. Func. Anal. 229 (2005), 317–374.
[21] S.L. Woronowicz: Quantum E(2)-group and its Pontryagin dual. Lett. Math. Phys. 23 (1991), 251–263.
[22] S.L. Woronowicz: Operator equalities related to quantum E(2) group. Commun. Math. Phys. 144 (1992),
417–428.
[23] S.L. Woronowicz: Quantum SU (2) and E(2) Groups. Contraction Procedure. Commun. Math. Phys. 149
(1992), 637–652.
[24] S.L. Woronowicz: Unbounded elements affiliated with C∗ -algebras and non-compact quantum groups. Com-
mun. Math. Phys. 136 (1991), 399–432.
[25] S.L. Woronowicz: C∗ -algebras generated by unbounded elements. Rev. Math. Phys. 7, no. 3 (1995), 481–521.
[26] S.L. Woronowicz: Compact quantum groups. In Symétries Quantiques, les Houches, Session LXIV 1995,
Elsevier 1998, pp. 845–884.
[27] S.L. Woronowicz: Quantum ‘az + b’ group on complex plane. Int. J. Math. 12, no. 4 (2001), 461–503.
[28] S.L. Woronowicz: Haar weight on some quantum groups. Group 24: Physical and mathematical aspects
of symmetries, Proceedings of the 24th International Colloquium on Group Theoretical Methods in Physics
Paris, 15–20 July 2002, Institute of Physics, Conference Series Number 173, pp. 763–772.
[29] S.L. Woronowicz & K. Napiórkowski: Operator theory in the C∗ -algebra framework. Rep. Math. Phys. 31
(1992), 353–371.

Institute of Mathematics, Polish Academy of Sciences

and
Department of Mathematical Methods in Physics, Faculty of Physics, University of Warsaw
E-mail address: [email protected]