This paper investigates the simplicity of $C^*$-algebras arising from $C^*$-symbolic dynamical systems, providing criteria and examples including irrational rotation Cuntz-Krieger algebras.
Contribution
It establishes conditions for simplicity of $C^*$-algebras from symbolic dynamical systems and explores specific examples like irrational rotation Cuntz-Krieger algebras.
Findings
01
Simplicity conditions for $C^*$-symbolic crossed products are identified.
02
Examples demonstrate the application of these conditions to specific algebras.
03
The study extends understanding of the structure of $C^*$-algebras from symbolic dynamics.
Abstract
We have introduced a notion of C∗-symbolic dynamical system in [K. Matsumoto: Actions of symbolic dynamical systems on C∗-algebras, to appear in J. Reine Angew. Math.], that is a finite family of endomorphisms of a C∗-algebra with some conditions. The endomorphisms are indexed by symbols and yield both a subshift and a C∗-algebra of a Hilbert C∗-bimodule. The associated C∗-algebra with the C∗-symbolic dynamical system is regarded as a crossed product by the subshift. We will study a simplicity condition of the C∗-algebras of the C∗-symbolic dynamical systems. Some examples such as irrational rotation Cuntz-Krieger algebras will be studied.
Equations176
A\CalG(i,α,j)={10 if there exists an edge e from vi to vj with λ(e)=α, otherwise
A\CalG(i,α,j)={10 if there exists an edge e from vi to vj with λ(e)=α, otherwise
ραL(Eil)=j=1∑m(l+1)Al,l+1(i,α,j)Ejl+1 for i=1,2,…,m(l),
ραL(Eil)=j=1∑m(l+1)Al,l+1(i,α,j)Ejl+1 for i=1,2,…,m(l),
No public reviews on file for this paper yet. If you reviewed it on a platform where reviews are public (OpenReview, ICLR, NeurIPS, ICML), you can paste yours below so the community can read it here.
Videos
No videos yet. Explain this paper in a talk, walkthrough, or lecture? Add one.
Taxonomy
TopicsAdvanced Operator Algebra Research · Advanced Topics in Algebra · Algebraic structures and combinatorial models
Full text
Actions of symbolic dynamical systems on C∗-algebras II.
Simplicity of C∗-symbolic crossed products and some examples
Kengo Matsumoto
Department of Mathematical Sciences
Yokohama City University
Seto 22-2, Kanazawa-ku, Yokohama 236-0027, JAPAN
We have introduced a notion of C∗-symbolic dynamical system
in [K. Matsumoto: Actions of symbolic dynamical systems on C∗-algebras,
to appear in J. Reine Angew. Math.],
that is a finite family of endomorphisms of a C∗-algebra
with some conditions.
The endomorphisms are indexed by symbols and yield both a subshift and a C∗-algebra of a Hilbert C∗-bimodule.
The associated C∗-algebra with the C∗-symbolic dynamical system
is regarded as a crossed product by the subshift.
We will study a simplicity condition of the
C∗-algebras of the C∗-symbolic dynamical systems.
Some examples
such as irrational rotation Cuntz-Krieger algebras will be studied.
In [CK],
J. Cuntz and W. Krieger have founded a close relationship
between symbolic dynamics and C∗-algebras (cf.[C], [C2]).
They constructed purely infinite simple C∗-algebras
from irreducible topological Markov shifts.
The C∗-algebras are called Cuntz-Krieger algebras.
In [Ma], the author introduced a notion of λ-graph system, whose matrix version is called symbolic matrix system.
A λ-graph system is a generalization of finite labeled graph and
presents a subshift.
He constructed C∗-algebras from λ-graph systems [Ma2]
as a generalization of the above Cuntz-Krieger algebras.
A λ-graph system gives rise to a finite family
{ρα}α∈Σ of endomorphisms of
a unital commutative AF-C∗-algbera \CalAL
with some conditions stated below.
A C∗-symbolic dynamical system, introduced in [Ma6], is a generalization of λ-graph system.
It is a finite family {ρα}α∈Σ
of endomorphisms of a unital C∗-algebra
\CalA
such that the closed ideal generated by ρα(1),α∈Σ
coincides with \CalA.
A finite labeled graph gives rise to a C∗-symbolic dynamical system
(\CalA,ρ,Σ) such that \CalA=CN for some
N∈N.
Conversely, if \CalA=CN,
the C∗-symbolic dynamical system comes from a finite labeled graph.
A λ-graph system L
gives rise to a C∗-symbolic dynamical system (\CalA,ρ,Σ)
such that \CalA is C(ΩL) for some compact Hausdorff space
ΩL with
dimΩL=0.
Conversely, if \CalA is C(X) for a compact Hausdorff space
X with
dimX=0,
the C∗-symbolic dynamical system comes from a λ-graph system.
A C∗-symbolic dynamical system (\CalA,ρ,Σ)
yields a nontrivial subshift
Λ(\CalA,ρ,Σ),
that we will denote by Λρ,
over Σ
and a Hilbert C∗-right \CalA-module
(ϕρ,\CalH\CalAρ,{uα}α∈Σ)
that has an orthogonal finite basis
{uα}α∈Σ
and a unital faithful diagonal left action
ϕρ:\CalA→L(\CalH\CalAρ).
It is called a Hilbert C∗-symbolic bimodule over\CalA,
and written as
(ϕρ,\CalH\CalAρ,{uα}α∈Σ).
By using general construction of C∗-algebras
from Hilbert C∗-bimodules established by M. Pimsner [Pim]
(cf. [Ka]),
the author has introduced
a C∗-algebra
denoted by
\CalA⋊ρΛ
from the Hilbert C∗-symbolic bimodule
(ϕρ,\CalH\CalAρ,{uα}α∈Σ),
where
Λ is the subshift Λρ
associated with
(\CalA,ρ,Σ).
We call the algebra
\CalA⋊ρΛtheC∗-symbolic crossed product of\CalAby the subshiftΛ.
If \CalA=C,
the subshift Λ is the full shift ΣZ,
and the C∗-algebra \CalA⋊ρΛ
is the Cuntz algebra \CalO∣Σ∣ of order ∣Σ∣.
If \CalA=C(X) with dimX=0,
there uniquely exists a λ-graph system L up to equivalence
such that the subshift Λ
is presented by L
and the C∗-algebra \CalA⋊ρΛ is the C∗-algebra \CalOL
associated with the λ-graph system L.
Conversely, for any subshift, that is presented by a λ-graph system L,
there exists a C∗-symbolic dynamical system
(\CalA,ρ,Σ) such that Λρ
is the subshift
presented by L,
the algebra \CalA is C(ΩL) with
dimΩL=0,
and the algebra \CalA⋊ρΛ is the C∗-algebra \CalOL
associated with L ([Ma6]).
If in particular, \CalA=Cn,
the subshift Λ is a sofic shift
and
\CalA⋊ρΛ is a Cuntz-Krieger algebra.
In this paper, a condition called (I) on (\CalA,ρ,Σ)
is introduced
as a generalization of condition (I) on the finite matrices of
Cuntz-Krieger [CK] and on the λ-graph systems [Ma2].
Under the assumption that (\CalA,ρ,Σ) satisfies condition (I),
the simplicity conditions of
the algebra \CalA⋊ρΛ is discussed in Section 3.
We further study ideal structure of \CalA⋊ρΛ
from the view point of quotients of the C∗-symbolic dynamical systems
in Section 4.
Related discussions have been studied in Kajiwara-Pinzari-Watatani’s paper [KPW] for the C∗-algebras of Hilbert C∗-bimodules
(cf. [Kat], [MS], [Tom], etc.).
They have studied simplicity condition and ideal structure of the C∗-algebras of Hilbert C∗-bimodules
in terms of the language of the Hilbert C∗-bimodules.
Our approach
to study the algebras \CalA⋊ρΛ is
from the view point of C∗-symbolic dynamical systems,
that is differnt from theirs.
In Section 5, we will study pure infiniteness of the algebras \CalA⋊ρΛ.
To obtain rich examples of the algebras \CalA⋊ρΛ,
we will in Section 6 construct
C∗-symbolic dynamical systems from a finite family of automorphisms
αi∈Aut(\CalB),i=1,…N on a unital C∗-algebra \CalB
and a C∗-symbolic dynamical systems (\CalA,ρ,Σ)
with
Σ={α1,…,αN}.
The C∗-symbolic dynamical system is denoted by
(\CalB⊗\CalA,ρΣ⊗,Σ) that is the tensor product between two C∗-symbolic dynamical systems(\CalB,α,Σ) and (\CalA,ρ,Σ).
As examples of C∗-symbolic crossed products,
continuous analogue of Cuntz-Krieger algebras
called irrational rotation Cuntz-Krieger algebras denoted by \CalO\CalG,θ1,…,θN
and
irrational rotation Cuntz algebras denoted by \CalOθ1,…,θN are studied in Sections 8 and 9.
They belongs to the class of the C∗-algebras of continuous graphs
by V. Deaconu ([De],[De2]).
The fixed point algebras \CalF\CalG,θ,…,θN of \CalO\CalG,θ1,…,θN under gauge actions are no longer AF-algebras.
They are AT-algebras.
In particular, the fixed point algebras \CalFθ1,…,θN
of \CalOθ1,…,θN under gauge actions
are simple AT-algebras of real rank zero with unique tracial state
if and only if differnce of rotation angles θi−θj is irrational for some i,j=1,…,N
(Theorem 9.4).
Throughout this paper, we denote by
Z+ and by N
the set of nonnegative integers and
the set of positive integers respectively.
A homomorphism and an isomorphism between C∗-algebras
mean a ∗-homomorphism and a ∗-isomorphism respectively.
An ideal of a C∗-algebra means a closed two sided ∗-ideal.
2. C∗-symbolic dynamical systems and their crossed products
Let \CalA be a unital C∗-algebra.
In what follows,
an endomorphism of \CalA means
a ∗-endomorphism of \CalA that does not necessarily preserve the unit
1\CalA of
\CalA.
The unit 1\CalA is denoted by 1 unless we specify.
We denote by End(\CalA) the set of all endomorphisms of \CalA.
Let Σ be a finite set.
A finite family of endomorphisms
ρα∈End(\CalA),α∈Σ
is said to be essential if
ρα(1)=0 for all α∈Σ
and
the closed ideal generated by ρα(1),α∈Σ
coincides with \CalA.
It is said to be faithful if for any nonzero x∈\CalA there exists a symbol α∈Σ such that ρα(x)=0.
We note that
{ρα}α∈Σ is faithful if and only if
the homomorphism
ξρ:a∈\CalA⟶[ρα(a)]α∈Σ∈⊕α∈Σ\CalA
is injective.
Definition ([Ma6]). A C∗-symbolic dynamical system is a triplet (\CalA,ρ,Σ)
consisting of a unital C∗-algebra \CalA
and
an essential and faithful finite family of endomorphisms ρα of \CalA indexed by α∈Σ.
Two C∗-symbolic dynamical systems (\CalA,ρ,Σ)
and
(\CalA′,ρ′,Σ′)
are said to be isomorphic if there exist
an isomorphism Φ:\CalA→\CalA′
and a bijection π:Σ→Σ′
such that
Φ∘ρα=ρπ(α)′∘Φ
for all
α∈Σ.
A C∗-symbolic dynamical system (\CalA,ρ,Σ)
yields a subshift Λ(\CalA,ρ,Σ)
over Σ such that a word α1⋯αk of Σ is admissible for Λ(\CalA,ρ,Σ)
if and only if
(ραk∘⋯∘ρα1)(1)=0
([Ma6;Proposition 2.1]).
The subshift Λ(\CalA,ρ,Σ)
will be denoted by
Λρ or simply by
Λ
in this paper.
Let \CalG=(G,λ) be a left-resolving finite labeled graph
with underlying finite directed graph G=(V,E) and labeling map
λ:E→Σ (see [LM; p.76]).
Denote by v1,…,vN the vertex set V.
Assume that every vertex has both an incoming edge and an outgoing edge.
Consider the N-dimensional commutative C∗-algebra
\CalA\CalG=CE1⊕⋯⊕CEN
where each minimal projection Ei
corresponds to the vertex vi
for
i=1,…,N.
Define an N×N-matrix for α∈Σ
by
[TABLE]
for i,j=1,…,N.
We set
ρα\CalG(Ei)=∑j=1NA\CalG(i,α,j)Ej
for
i=1,…,N,α∈Σ.
Then
ρα\CalG,α∈Σ
define endomorphisms of \CalA\CalG
such that
(\CalA\CalG,ρ\CalG,Σ)
is a C∗-symbolic dynamical system
such that the algebra \CalA\CalG is CN,
and the subshift
Λρ\CalG
is the sofic shift
Λ\CalG
presented by \CalG.
Conversely,
for a C∗-symbolic dynamical system
(\CalA,ρ,Σ),
if \CalA is CN,
there exists a left-resolving labeled graph
\CalG such that \CalA=\CalA\CalG
and
Λρ=Λ\CalG
the sofic shift
presented by \CalG ([Ma6;Proposition 2.2]).
More generally let L be a λ-graph system
(V,E,λ,ι) over Σ (see [Ma]).
Its vertex set V is ∪l=0∞Vl.
We equip Vl with discrete topology.
We denote by
ΩL
the
compact Hausdorff space
with dimΩL=0
of the projective limit
V0→ι←V1→ι←V2→ι←⋯,
as in [Ma2;Section 2].
The algebra C(Vl) of all continuous functions on Vl,
denoted by \CalAL,l,
is the direct sum
\CalAL,l=CE1l⊕⋯⊕CEm(l)l
where
each
minimal projection
Eil
corresponds to
the vertex
vil
for
i=1,…,m(l).
Let \CalAL be the commutative C∗-algebra
C(ΩL)=liml→∞{ι∗:\CalAL,l→\CalAL,l+1}.
Let Al,l+1,l∈Z+ be the matrices defined in [Ma2; Theorem A].
For a symbol α∈Σ we set
[TABLE]
so that
ραL defines an endomorphism of \CalAL.
We have a
C∗-symbolic dynamical system
(\CalAL,ρL,Σ)
such that
the C∗-algebra \CalAL is C(ΩL)
with dimΩL=0,
and the subshift
ΛρL
coincides with the subshift ΛL presented by L.
Conversely, for a
C∗-symbolic dynamical system
(\CalA,ρ,Σ),
if the algebra \CalA is C(X) with dimX=0,
there exists a λ-graph system L over Σ
such that the associated C∗-symbolic dynamical system
(\CalAL,ρL,Σ)
is isomorphic to
(\CalA,ρ,Σ) ([Ma6;Theorem 2.4]).
Let L and L′
be predecessor-separated λ-graph systems over
Σ and Σ′ respectively.
Then
(\CalAL,ρL,Σ) is isomorphic to
(\CalAL′,ρL′,Σ′) if and only if
L and L′ are equivalent.
In this case, the presented subshifts ΛL
and ΛL′ are identified through a symbolic conjugacy.
Hence the equivalence classes of the λ-graph systems are identified with the isomorphism classes of the C∗-symbolic dynamical systems of the commutative AF-algebras.
We say that a subshiftΛacts on aC∗-algebra\CalA if there exists a C∗-symbolic dynamical system
(\CalA,ρ,Σ)
such that the associated subshift Λρ
is Λ.
For a C∗-symbolic dynamical system
(\CalA,ρ,Σ),
we have a Hilbert C∗-bimodule
(ϕρ,\CalH\CalAρ,{uα}α∈Σ)
called a Hilbert C∗-symbolic bimodule ([Ma6]).
We then have a C∗-algebra
by using the Pimsner’s general construction of
C∗-algebras from Hilbert C∗-bimodules
[Pim]
(cf. [Ka], see also [KPW], [KW], [Kat], [MS], [PWY], [Sch] etc.).
We denote the C∗-algebra by \CalA⋊ρΛ, where
Λ is the subshift Λρ
associated with
(\CalA,ρ,Σ).
We call the algebra
\CalA⋊ρΛtheC∗-symbolic crossed product of\CalAby the subshiftΛ.
Proposition 2.1([Ma6;Proposition 4.1])
The C∗-symbolic crossed product \CalA⋊ρΛ
is the universal C∗-algebra
C∗(x,Sα;x∈\CalA,α∈Σ)
generated by
x∈\CalA
and partial isometries Sα,α∈Σ
subject to the following relations called (ρ):
[TABLE]
for all x∈\CalA and α∈Σ.
Furthermore for α1,…,αk∈Σ,
a word (α1,…,αk) is admissible for the subshift
Λ
if and only if
Sα1⋯Sαk=0.
Assume that \CalA is commutative.
Then we know ([Ma6;Theorem 4.2])
3. Condition (I) for C∗-symbolic dynamical systems
The notion of condition (I) for finite square matrices
with entries in {0,1} has been introduced in [CK].
The condition
gives rise to the uniqueness of the associated Cuntz-Krieger algebras
under the canonical relations of the generating partial isometries.
The condition has been generalized by many authors to
corresponding conditions for generalizations of the Cuntz-Krieger algebras,
for instance, infinite directed graphs ([KPRR]), infinite matrices with entries in {0,1} ([EL]), Hilbert C∗-bimodules ([KPW]), etc.
(see also [Re], [Ka2],[Tom2],etc.).
The condition (I) for λ-graph systems has been also defined in [Ma2]
to prove the uniqueness of the C∗-algebra \CalOL
under the canonical relations of generators.
In this section,
we will introduce the notion of condition (I)
for C∗-symbolic dynamical systems
to prove
the uniqueness of the C∗-algebras
\CalA⋊ρΛ
under the relation (ρ).
In [KPW], a condition called (I)-free has been introduced.
The condition is similar condition to our condition (I).
The discussions given in [KPW]
is also similar ones to ours in this section.
We will give
complete descriptions in our discussions for the sake of completeness.
Throughout this paper, for a subset F of a C∗-algebra \CalB,
we denote by C∗(F) the C∗-subalgebra of \CalB generated by F.
In what follows, (\CalA,ρ,Σ) denotes a C∗-symbolic dynamical system
and Λ the associated subshift Λρ.
We denote by Λk the set of admissible words μ of
Λ with length ∣μ∣=k.
Put
Λ∗=∪k=0∞Λk,
where Λ0 denotes the empty word.
Let Sα,α∈Σ
be the partial isometries in \CalA⋊ρΛ satisfying the relation (ρ)
in Proposition 2.1.
For
μ=(μ1,…,μk)∈Λk,
we put
Sμ=Sμ1⋯Sμk
and
ρμ=ρμk∘⋯∘ρμ1.
In the algebra \CalA⋊ρΛ,
we set
[TABLE]
The identity
SμxSν∗=∑α∈ΣSμαρα(x)Sνα∗
for x∈\CalA and μ,ν∈Λk holds
so that the algebra
\CalFρk is embedded into the algebra
\CalFρk+1
such that
∪k∈Z+\CalFρk is dense in \CalFρ.
The gauge action ρ^
of the circle group T={z∈C∣∣z∣=1}
on \CalA⋊ρΛ is defined by
ρ^z(x)=x for x∈\CalA
and ρ^z(Sα)=zSα
for α∈Σ.
The fixed point algebra of \CalA⋊ρΛ under ρ^ is denoted by
(\CalA⋊ρΛ)ρ^.
Let \CalEρ:\CalA⋊ρΛ⟶(\CalA⋊ρΛ)ρ^ be the conditional expectaton defined by
[TABLE]
It is routine to check that
(\CalA⋊ρΛ)ρ^=\CalFρ.
Let \CalB be a unital C∗-algebra.
Suppose that there exist an injective homomorphism
π:\CalA⟶\CalB preserving their units
and
a family sα∈\CalB,α∈Σ
of partial isometries
satisfying
[TABLE]
for all x∈\CalA and α∈Σ.
Put
\CalA=π(\CalA)⊂
and
ρ~α(π(x))=π(ρα(x)),x∈\CalA.
We then have
Lemma 3.1
The triple
(\CalA,ρ~,Σ) is a C∗-symbolic dynamical system
such that the presented subshift
Λρ~
is the same as the one Λ(=Λρ)
presented by
(\CalA,ρ,Σ).
Let
\CalOπ,s
be the C∗-subalgebra of \CalB generated by
π(x) and sα for x∈\CalA,α∈Σ.
In the algebra \CalOπ,s, we set
[TABLE]
By the universality of the algebra \CalA⋊ρΛ,
the correspondence
[TABLE]
extends to a surjective homomorphism
π~:\CalA⋊ρΛ⟶\CalOπ,s.
Lemma 3.2
The restriction of π~ to the subalgebra
\CalFρ
is an isomorphism from
\CalFρ to \CalFπ,s.
Demonstration Proof
It suffices to show that π~ is injective on
\CalFρk.
Suppose that
∑μ,ν∈Λksμπ(xμ,ν)sν∗=0
for
∑μ,ν∈ΛkSμxμ,νSν∗∈\CalFρ
with
xμ,ν∈\CalA.
For ξ,η∈Λk, it follows that
[TABLE]
As π:\CalA⟶\CalB is injective,
one has
ρξ(1)xξ,ηρη(1)=0
so that
Sξxξ,ηSη∗=0.
This implies that
∑μ,ν∈ΛkSμxμ,νSν∗=0.
∎
**Definition. **
A C∗-symbolic dynamical system (\CalA,ρ,Σ)
satisfies *condition *(I) if
there exists a unital increasing sequence
[TABLE]
of C∗-subalgebras of \CalA such that
ρα(\CalAl)⊂\CalAl+1
for all l∈Z+,α∈Σ
and
the union ∪l∈Z+\CalAl is dense in \CalA
and
for k,l∈N with k≤l,
there exists a projection
qkl∈\CalDρ∩\CalAl′(={x∈\CalDρ∣xa=ax for a∈\CalAl})
such that
As the projection qkl belongs to
the diagonal subalgebra \CalDρ of \CalFρ,
the condition (I) of (\CalA,ρ,Σ)
is intrinsically determined by (\CalA,ρ,Σ)
by virtue of Lemma 3.2.
If a λ-graph system L over Σ
satisfies condition (I),
then (\CalAL,ρL,Σ)
satisfies condition (I) (cf. [Ma2;lemma 4.1]).
We now assume that (\CalA,ρ,Σ) satisfies condition (I).
We set for k≤l
[TABLE]
There exists an inclusion relation
\CalFlk⊂\CalFl′k′ for k≤k′ and l≤l′.
We put a projection
Qkl=ϕρk(qkl) in \CalDρ.
Lemma 3.3
The map X∈\CalFρ,lk⟶QklXQkl∈Qkl\CalFρ,lkQkl
is a surjective isomorphism.
Demonstration Proof
As qkl commutes with \CalAl,
for x∈\CalAl and μ,ν∈Λk, we have
[TABLE]
and similarly
SμxSν∗Qkl=SμxqklSν∗
so that Qkl commutes with SμxSν∗.
Hence the map
X∈\CalFρ,lk⟶QklXQkl∈Qkl\CalFρ,lkQkl
defines a surjective homomorphism.
It remains to show that it is injective.
Suppose that
Qkl(∑μ,ν∈ΛkSμxμ,νSν∗)Qkl=0
for
X=∑μ,ν∈ΛkSμxμ,νSν∗
with
xμ,ν∈\CalAl.
For ξ,η∈Λk, one has
[TABLE]
so that
0=Sξ∗QklSξxξ,ηSη∗Sη=Sξ∗Sξqklρξ(1)xξ,ηSη∗Sη=qklρξ(1)xξ,ηρη(1).
Hence
ρξ(1)xξ,ηρη(1)=0
by condition (I).
Thus
Sξxξ,ηSη∗=0,
so that
∑ξ,η∈ΛkSξxξ,ηSη∗=0.
∎
Lemma 3.4
QklSμQkl=0 for μ∈Λ∗ with ∣μ∣≤k≤l.
Demonstration Proof
By condition (I), we have
Qklϕρm(Qkl)=0 for 1≤m≤k.
For μ∈Λ∗ with ∣μ∣≤k, one has
ϕρ∣μ∣(Qkl)Sμ=SμQklSμ∗Sμ=SμQkl.
Hence we have
0=Qklϕρ∣μ∣(Qkl)Sμ=QklSμQkl.
∎
As a result, we have
Lemma 3.5
The projections Qkl in \CalDρ
satisfy the following conditions:
We note that
QklSμQkl=0 if and only if QklSμQklSμ∗=0.
Since
QklSμQklSμ∗ belongs to the algebra
\CalFρ,
the condition
QklSμQkl=0
is determined in the algebraic structure of
\CalFρ.
As the restriction of
π~:\CalA⋊ρΛ⟶\CalOπ,s
to
\CalFρ
yields an isomorphism onto
\CalFπ,s,
by putting
Qkl=π~(Qkl)
we have
Lemma 3.6
The projections Qkl
in \CalDπ,s satisfy the following conditions:
Proposition 3.7
There exists a conditional expectation
\CalEπ,s:\CalOπ,s⟶\CalFπ,s
such that
\CalEπ,s∘π~=π~∘\CalEρ.
Demonstration Proof
Let \CalPπ,s
be the ∗-subalgebra of \CalOπ,s
generated algebraically by
π(x),sα for x∈\CalA,α∈Σ.
Then any
X∈\CalPπ,s can be written as a finite sum
[TABLE]
Thanks to the previous lemma and a usual argument of [CK],
the element
X0∈\CalFπ,s
is unique for X∈\CalPπ,s
and the inequality
∥X0∥≤∥X∥
holds.
The map
X∈\CalPπ,s→X0∈\CalFπ,s
can be extended to the desired expectation
\CalEπ,s:\CalOπ,s⟶\CalFπ,s.
∎
Therefore we have
Theorem 3.8
Assume that (\CalA,ρ,Σ)
satisfies condition (I).
The homomorphism π~:\CalA⋊ρΛ⟶\CalOπ,s
defined by
[TABLE]
becomes a surjective isomorphism, and hence
the C∗-algebras \CalA⋊ρΛ and \CalOπ,s
are canonically isomorphic through π~.
Demonstration Proof
The map π~:\CalFρ→\CalFπ,s
is isomorphic and satisfies
\CalEπ,s∘π~=π~∘\CalEρ.
Since
\CalEρ:\CalA⋊ρΛ⟶\CalFρ
is faithful,
a routine argument shows that the homomorphism
π~:\CalA⋊ρΛ⟶\CalOπ,s
is actually an isomorphism.
∎
Hence the following uniqueness of the C∗-algebra \CalA⋊ρΛ holds.
Theorem 3.9
Assume that (\CalA,ρ,Σ)
satisfies condition (I).
The C∗-algebra \CalA⋊ρΛ is the unique C∗-algebra subject to the relation
(ρ).
This means that if there exist a unital C∗-algebra \CalB and an
injective homomorphism π:\CalA⟶\CalB and a family
sα∈\CalB,α∈Σ
of nonzero partial isometries satisfying the folloing relations:
[TABLE]
for all x∈\CalA and α∈Σ,
then the correspondence
[TABLE]
extends to an isomorphism
π~ from \CalA⋊ρΛ onto the C∗-subalgebra
\CalOπ,s of \CalB
generated by π(x),x∈\CalA and sα,α∈Σ.
As a corollary we have
Corollary 3.10
Assume that (\CalA,ρ,Σ)
satisfies condition (I).
For any nontrivial ideal \CalI
of \CalA⋊ρΛ, one has
\CalI∩\CalA={0}.
Demonstration Proof
Suppose that
\CalI∩\CalA={0}.
Hence Sα∈\CalI
for all α∈Σ.
By Theorem 3.9, the quotient map
q:\CalA⋊ρΛ⟶\CalA⋊ρΛ/\CalI
must be injective so that \CalI is trivial.
∎
Let
λρ:\CalA→\CalA
be the completely positive map on \CalA defined by
λρ(x)=∑α∈Σρα(x)
for x∈\CalA.
Definition.(\CalA,ρ,Σ)
is said to be *irreduible * if
there exists no nontrivial ideal of \CalA invariant under λρ.
Therefore we have
Corollary 3.11
Assume that (\CalA,ρ,Σ)
satisfies condition (I).
If (\CalA,ρ,Σ) is irreducible,
the C∗-algebra \CalA⋊ρΛ is simple.
4. Quotients of C∗-symbolic dynamical systems
In this section,
we will study ideal structure of the C∗-symbolic crossed products
\CalA⋊ρΛ, related to quotients of C∗-symbolic dynamical systems.
The ideal structure of C∗-algebras of Hilbert C∗-bimodules
has been studied in Kajiwara, Pinzari and Watatani’s paper [KPW]
(cf. [Kat3]).
Their paper is written in the language of Hilbert C∗-bimodules.
In this section we will directly study ideal structure of
the C∗-symbolic crossed products
\CalA⋊ρΛ by using the language of C∗-symbolic dynamical systems.
We fix a C∗-symbolic dynamical system (\CalA,ρ,Σ).
An ideal J of \CalA is said to be ρ-*invariant * if
ρα(J)⊂J for all α∈Σ.
It is said to be *saturated * if ρα(x)∈J for all α∈Σ implies x∈J.
Lemma 4.1
Let J be an ideal of \CalA.
Demonstration Proof
(i) Suppose that J satisfies λρ(J)⊂J.
For x∈J one has
λρ(x∗x)≥ρα(x∗x)=ρα(x)∗ρα(x)
so that
ρα(x)∗ρα(x)∈J
because ideal is hereditary.
Hence ρα(x) belongs to J.
The only if part is clear.
(ii) Suppose that J is saturated and λρ(a)∈J for
0≤a∈\CalA.
Since λρ(a)≥ρα(a) and J is hereditary,
one has a∈J.
Conversely suppose that x∈\CalA satisfies
ρα(x)∈J for all α∈Σ.
As
λρ(x∗x)=∑α∈Σρα(x)∗ρα(x),λρ(x∗x) belongs to J.
Hence the condition of the if part implies that x∗x∈J
so that x∈J.
∎
Let J be a ρ-invariant saturated ideal of \CalA.
We denote by \CalIJ the ideal of \CalA⋊ρΛ generated by J.
Lemma 4.2
The ideal \CalIJ
is the closure of linear combinations of elements of the form
Sμcμ,νSν∗
for
cμ,ν∈J.
Demonstration Proof
Elements x and y of \CalA⋊ρΛ
are approximated by finite sums of elements of the form
Sμaμ,νSν∗
and
Sξbξ,ηSη∗
for
aμ,ν,bξ,η∈\CalA
respectively.
Hence xcy is approximated by elemnts of the form
[TABLE]
In case of ∣ν∣≥∣ξ∣,
one has ν=νˉν′ with ∣νˉ∣=∣ξ∣
so that
[TABLE]
Hence
Sμaμ,νSν∗cSξbξ,ηSη∗
is
Sμaμ,νρν′(ρνˉ(c)bξ,η)Sην′∗
or zero.
Since J is ρ-invariant,
it is of the form
Sμcμ,νSην′∗ for some cμ,ν∈J
or zero.
The argument in case of ∣ν∣≤∣ξ∣ is similar.
Since the ideal
\CalIJ
is the closure of elements of the form
∑i=1nxiciyi
for
xi,yi∈\CalA⋊ρΛ
and
ci∈J,
the assertion is proved.
∎
We set
[TABLE]
Lemma 4.3
Demonstration Proof
(i)
Since the elements of the finite sum
∑μSμcμSμ∗ for cμ∈J
are contained in \CalIJ∩\CalDρ,
the inclusion relation
\CalDJ⊂\CalIJ∩\CalDρ
is clear.
Let
\CalIJalg and \CalDJalg
be the algebraic linear spans of
Sμcμ,νSν∗ for cμ,ν∈J
and
SμcμSμ∗ for cμ∈J
respectively.
For any x∈\CalIJ∩\CalDρ
take xn∈\CalIJalg
such that
∥xn−x∥→0.
Let
\CalEρ:\CalA⋊ρΛ→\CalFρ
be the conditional expectation defined previously,
and \CalED:\CalFρ→\CalDρ
the conditional expectation defined by taking diagonal elements.
The composition
\CalE\CalDρ=\CalED∘\CalEρ
is the conditional expectation from \CalA⋊ρΛ to \CalDρ
that satisfies
\CalE\CalDρ(\CalIJalg)=\CalDJalg.
Since
\CalE\CalDρ(x)=x and
the inequality
∥x−\CalE\CalDρ(xn)∥≤∥x−xn∥
holds,
x belongs to the closure \CalDJ
of \CalDJalg.
Hence we have
\CalIJ∩\CalDρ⊂\CalDJ
so that
\CalDJ=\CalIJ∩\CalDρ.
As \CalA is a subalgebra of \CalDρ,
the equality
\CalDJ∩\CalA=\CalIJ∩\CalA
holds.
(ii)
An element
x∈\CalDJk is of the form
∑∣μ∣≤kSμcμSμ∗ for cμ∈J.
As
SνcνSν∗=∑α∈ΣSναρα(cν)Sνα∗
and J is ρ-invariant,
x can be written as
x=∑∣ν∣=kSνcνSν∗ for cν∈J,
and the element
λρk(x)=∑∣ν∣=kρν(1)cνρν(1)
belongs to J.
Further suppose that x is an element of \CalA.
Since J is saturated, by Lemma 4.1, one has
x∈J.
Hence
the inclusion relation
\CalA∩\CalDJk⊂J holds.
The converse inclusion relation is clear so that
\CalA∩\CalDJk=J.
∎
Lemm 4.4
\CalA∩\CalDJ=J.
Demonstration Proof
Since the inclusion relation
\CalA∩\CalDJ⊃J is clear,
there exists a natural surjective homomorphism
from
\CalA/J onto \CalA/\CalA∩\CalDJ.
For an element a of a C∗-algebra \CalB,
we denote by ∥[a]\CalB/I∥
the norm of the quotient image
[a]\CalB/I of a in the quotient
\CalB/I of \CalB by an ideal I.
As the inclusion \CalA↪\CalDρ
induces the inclusions both
\CalA/\CalA∩\CalDJ↪\CalDρ/\CalDJ
and
\CalA/\CalA∩\CalDJk↪\CalDρ/\CalDJk,
one has for a∈\CalA
[TABLE]
Note that \CalDJ is the inductive limit of
\CalDJk,k=0,1,….
It then follows that
[TABLE]
and hence
∥[a]\CalA/\CalA∩\CalDJ∥=∥[a]\CalA/J∥
by Lemma 4.3 (ii).
Thus the quotient map
\CalA/J→\CalA/\CalA∩\CalDJ
is isometric so that
\CalA∩\CalDJ=J.
∎
By Lemma 4.3 and Lemma 4.4, one has
Proposition 4.5
\CalIJ∩\CalA=J.
We will now consider quotient C∗-symbolic dynamical systems.
Let J be a ρ-invariant saturated ideal of \CalA.
We set
ΣJ={α∈Σ∣ρα(1)∈J}.
We denote by [x] the class of x∈\CalA in the quotient \CalA/J.
Put
[TABLE]
As J is ρ-invariant and saturated,
ραJ is well-defined
and
the family
{ραJ}α∈ΣJ
is a faithful and essential endomorphisms of \CalA/J.
We call the C∗-symbolic dynamical system (\CalA/J,ρJ,ΣJ)
the quotient of (\CalA,ρ,Σ) by the ideal J.
We denote by ΛJ the associated subshift for the quotient
(\CalA/J,ρJ,ΣJ).
Definition.
A C∗-symbolic dynamical system
(\CalA,ρ,Σ) is said to satisfy *condition * (II)
if for any proper ρ-invariant saturated ideal J of \CalA,
the quotient C∗-symbolic dynamical system (\CalA/J,ρJ,ΣJ)
satisfies condition (I).
Let \CalI be a proper ideal of \CalA⋊ρΛ.
Put J\CalI:=\CalI∩\CalA.
Lemma 4.6
Demonstration Proof
(i) By condition (I), J\CalI is a nozero ideal of \CalA,
that is ρ-invariant.
If ρα(x) belongs to J\CalI for all α∈Σ,
the identity
x=∑α∈ΣSαρα(x)Sα∗
implies x∈\CalI, so that J\CalI is saturated.
The equality J\CalIJ=J follows from Proposition 4.5.
(ii)
Let π\CalI:\CalA⋊ρΛ→\CalA⋊ρΛ/\CalI
be the quotient map.
Put sα=π\CalI(Sα).
Then α∈ΣJ\CalI if and only if sα=0.
The following relations
[TABLE]
for x∈\CalA,α∈ΣJ\CalI
hold.
As
(\CalA/J\CalI,ρJ\CalI,ΣJ\CalI)
satisfies condition (I),
the uniqueness of the C∗-symbolic crossed product
(\CalA/J\CalI)⋊ρJ\CalIΛJ\CalI
yields a canonical isomorphism to the quotient algebra
\CalA⋊ρΛ/\CalI.
∎
Let \CalIJ\CalI be the ideal of \CalA⋊ρΛ generated by J\CalI.
Since J\CalI⊂\CalI, the inclusion relation
\CalIJ\CalI⊂\CalI is clear.
Lemma 4.7
If (\CalA,ρ,Σ) satisfies condition (II),
then there exists a canonical isomorphism from
(\CalA/J\CalI)⋊ρJ\CalIΛJ\CalI
to the quotient algebra
\CalA⋊ρΛ/\CalIJ\CalI.
Demonstration Proof
Take an arbitrary element x∈\CalA.
If x∈J\CalI, then x∈\CalIJ\CalI.
Conversely x∈\CalIJ\CalI implies x∈J\CalI by Proposition 4.5.
Hence
x∈J\CalI if and only if x∈\CalIJ\CalI.
For α∈Σ,
we have
Sα∈\CalIJ\CalI
if and only if
Sα∗Sα∈\CalIJ\CalI∩\CalA.
By Proposition 4.5, the latter condition is equivalent to the condition
ρα(1)∈J\CalI.
We know that
α∈ΣJ\CalI
if and only if Sα∈\CalIJ\CalI.
By the uniqueness of the algebra
(\CalA/J\CalI)⋊ρJ\CalIΛJ\CalI,
it is canonically isomorphic to the quotient algebra
\CalA⋊ρΛ/\CalIJ\CalI.
∎
Proposition 4.8
Suppose that (\CalA,ρ,Σ) satisfies condition (II).
For a proper ideal \CalI of \CalA⋊ρΛ,
let \CalIJ\CalI be the ideal of \CalA⋊ρΛ generated by
J\CalI.
Then we have
\CalIJ\CalI=\CalI.
Demonstration Proof
Since \CalIJ\CalI⊂\CalI,
there exists a quotient map
q\CalI:\CalA⋊ρΛ/\CalIJ\CalI→\CalA⋊ρΛ/\CalI.
By Lemma 4.6, and Lemma 4.7, there exist canonical isomorphisms
[TABLE]
Since q\CalI=π1∘π2−1,
it is ismorphism so that we have \CalIJ\CalI=\CalI.
∎
Therefore we have
Theorem 4.9
Suppose that (\CalA,ρ,Σ) satisfies condition (II).
There exists an inclusion preserving bijective correspondence between
ρ-invariant saturated ideals of \CalA and ideals of \CalA⋊ρΛ,
through the correspondences:
J→\CalIJ and J\CalI←\CalI.
5. Pure infiniteness
In this section we will show that
the C∗-symbolic crossed product \CalA⋊ρΛ
is purely infinite
if (\CalA,ρ,Σ) satisfies some conditions.
Definition.
A C∗-symbolic dynamical system
(\CalA,ρ,Σ) is said to be central if the projections
{ρμ(1)∣μ∈Λ∗} contained
in the center Z\CalA of \CalA.
It is said to be commutative
if \CalA is commutative.
Hence if (\CalA,ρ,Σ) is central,
the inequality ∑α∈Σρα(1)≥1 holds.
Let \CalAρ be the C∗-subalgebra of \CalA generated by the projections
ρμ(1),μ∈Λ∗.
Lemma 5.1
Assume that
(\CalA,ρ,Σ) is central.
Then
there exists a λ-graph system
Lρ
over Σ such that
the presented subshift
ΛLρ
coincides with the subshift Λ
presented by (\CalA,ρ,Σ),
and there exists a unital embedding of
\CalOLρ
into \CalA⋊ρΛ.
Demonstration Proof
Put
\CalAρ,0=C.
For l∈Z+, we define the C∗-algebra
\CalAρ,l+1
to be the C∗-subalgebra of \CalA
generated by the elements
ρα(x) for α∈Σ,x∈\CalAρ,l.
Hence the C∗-algebra \CalAρ is generated by
∪l=0∞\CalAρ,l.
Then
(\CalAρ,ρ,Σ) is a C∗-symbolic dynamical system
such that
\CalAρ is commutative and AF,
so that
there exists a λ-graph system
Lρ
over Σ such that
\CalAρ=\CalALρ.
The presented subshift
ΛLρ
coincides with the subshift Λ.
It is easy to see that there exists a unital embedding of
\CalOLρ
into \CalA⋊ρΛ by their universalities.
∎
In the rest of this section we assume that (\CalA,ρ,Σ) satisfies condition (I).
Definition.(\CalA,ρ,Σ) is said to be effective
if for l∈Z+ and a nonzero positive element a∈\CalAl,
there exist K∈N and a nonzero positive element b∈\CalAρ
such that
[TABLE]
where \CalAl is a C∗-subalgebra of \CalA appearing in the definition of condition (I).
In what follows, we assume that (\CalA,ρ,Σ) is effective,
and central.
Let L=Lρ be the λ-graph system associated to (\CalA,ρ,Σ) as in Lemma 5.1.
We further assume that the algebra
\CalOL is simple, purely infinite.
In [Ma2], [Ma3], conditions that the algebra \CalOL becomes simple, purely infinite is studied.
Lemma 5.2
For k≤l∈Z+
and a nonzero positive element
a∈\CalFρ,lk,
there exists an element V∈\CalA⋊ρΛ
such that
VaV∗=1.
Demonstration Proof
An element
a∈\CalFρ,lk is of the form
a=∑μ,ν∈ΛkSμaμ,νSν∗
for some aμ,ν∈\CalAl such that
Sμ∗aSν=aμ,ν.
Since a is a nonzero positive element,
there exists ξ∈Λk
such that
Sξ∗aSξ(=aξ,ξ)=0.
As we are assuming that
(\CalA,ρ,Σ) is effective,
there exists K∈N and a nonzero positive element b∈\CalAρ
[TABLE]
Put T=∑μ∈ΛKSμ∈\CalA⋊ρΛ.
One has T∗Sξ∗aSξT≥b.
Now b∈\CalAρ⊂\CalOL
and \CalOL is simple, purely infinite.
We may find V0∈\CalOL
such that V0bVo∗=1
so that
V0T∗Sξ∗aSξTV0∗≥1.
Hence there exists V∈\CalA⋊ρΛ such that
VaV∗=1.
∎
Lemma 5.3
Keep the above situation.
We may take V∈\CalA⋊ρΛ in the preceding lemma such as
VaV∗=1
and
∥V∥<∥a∥−21+ϵ
for a given
ϵ>0.
Demonstration Proof
We may assume that ∥a∥=1
and there exists p∈Sp(a) such that 0<p<1.
Take 0<ϵ<21
such that ϵ<1−p.
Define a function f∈C([0,1])
by setting
[TABLE]
Put b=f(a), that is not invertible.
By Lemma 5.2, there exists V∈\CalA⋊ρΛ
such that
VbV∗=1.
We set S=b21V∗
and
P=SS∗.
Then S is a proper isometry
such that P≤∥V∥b.
As P≤Ea([1−ϵ,1]),
Ea([1−ϵ,1]) is the spectral measure of a
for the interval [1−ϵ,1],
one has
PaP≥(1−ϵ)P.
Put
D=S∗aS so that
D≥S∗(1−ϵ)PS=(1−ϵ)1.
Hence D is invertible.
Set
V1=D−21S∗.
Then one sees that
V1aV1∗=1
and
∥V1∥<(1−ϵ)−21<1+ϵ.
∎
Let \CalEρ:\CalA⋊ρΛ→\CalFρ be
the conditional expectation defined in Section 3.
Lemma 5.4
For a nonzero X∈\CalA⋊ρΛ and ϵ>0,
there exists a projection Q∈\CalDρ
and a nonzero positive element Z∈\CalFρ,kl for some k≤l
such that
[TABLE]
Demonstration Proof
We may assume that
∥\CalEρ(X∗X)∥=1.
Let
\CalPρ be the
∗-algebra generated algebraically by
Sα,α∈Σ and
x∈\CalA.
For any 0<ϵ<41,
find
0≤Y∈\CalPρ
such that
∥X∗X−Y∥<2ϵ
so that
∥\CalEρ(Y)∥>1−2ϵ.
As in the discussion in [Ma3;Section 3], the element
Y is expressed
as
[TABLE]
Take k≤l large enough such that
Y−ν,Y0,Yμ∈\CalFρ,kl
for all
μ,ν
in the above expression.
Now
(\CalA,ρ,Σ) satisfies condition (I).
Take a sequence Qkl∈\CalDρ
of projections as in Section 3.
As \CalEρ(Y)=Y0
and Qkl commutes with \CalFρ,kl,
it follows that by Lemma 3.5 (c),
QklYQkl=Qkl\CalEρ(Y)Qkl.
Since Qkl\CalEρ(Y)Qkl∈\CalFρ,
there exists 0≤Z∈\CalFρ,k′l′
for some k′≤l′
such that
∥Qkl\CalEρ(Y)Qkl−Z∥<2ϵ.
By Lemma 3.3,
we note
∥Qkl\CalEρ(Y)Qkl∥=∥\CalEρ(Y)∥ so that
[TABLE]
and
[TABLE]
∎
Therefore we have
Theorem 5.5
Assume that (\CalA,ρ,Σ) is central, irreducible and satisfies condition (I).
Let L be the associated λ-graph system to (\CalA,ρ,Σ).
If (\CalA,ρ,Σ) is effective and \CalOL is simple, purely infinite, then
\CalA⋊ρΛ is simple, purely infinite.
Demonstration Proof
It suffices to show that
for any nonzero X∈\CalA⋊ρΛ, ther exist
A,B∈\CalA⋊ρΛ such that
AXB=1.
By the previous lemma there exists
a projection Q∈\CalDρ
and a nonzero positive element Z∈\CalFρ,kl
for some k≤l such that
∥QX∗XQ−Z∥<ϵ.
We may assume that ∥\CalEρ(X∗X)∥=1
so that
1−ϵ<∥Z∥<1+ϵ.
By Lemma 5.3, take an element
V∈\CalA⋊ρΛ such that
[TABLE]
It follows that
[TABLE]
We may take ϵ>0 small enough so that
∥VQX∗XQV∗−1∥<1
and hence
VQX∗XQV∗ is invertible in \CalA⋊ρΛ.
Thus we complete the proof.
∎
6. Tensor products of C∗-symbolic dynamical systems
In this section, we will consider tensor products between
C∗-symbolic dynamical systems and finite families of automorphisms of unital C∗-algebras.
This construction yields
interesting examples of C∗-symbolic dynamical systems
beyond λ-graph systems,
that will be studied in the following sections.
Throughout this section,
we fix a unital C∗-algebra \CalB
and a finite family of automorphisms
αi∈Aut(\CalB),i=1,…,N of \CalB.
Tensor products ⊗ between C∗-algebras always mean
the minimal C∗-tensor products ⊗min.
We set Σ={α1,…,αN}.
Consider a C∗-symbolic dynamical system
(\CalA,ρ,Σ).
Proposition 6.1
For αi∈Σ,i=1,…,N,
define ραiΣ⊗∈End(\CalB⊗\CalA)
by setting
[TABLE]
Then
(\CalB⊗\CalA,ρΣ⊗,Σ)
becomes a C∗-symbolic dynamical system over Σ such that the presented subshift ΛρΣ⊗
is the same as the subshift Λρ
presented by (\CalA,ρ,Σ).
Demonstration Proof
We will first prove that
(\CalB⊗\CalA,ρΣ⊗,Σ)
is a C∗-symbolic dynamical system.
Since
{ραi}i=1N is essential,
for ϵ>0,
there exist
xi,j,yi,j∈\CalA,j=1,…,n(i),i=1,…,N
such that
[TABLE]
so that we have
[TABLE]
Hence the closed ideal generated by
{ραiΣ⊗(1):i=1,…,N}
is all of \CalB⊗\CalA,
so that
{ραiΣ⊗}i=1N
is essential.
Since {ραi}i=1N is faithful on \CalA,
the homomorphism
ξρ:\CalA⟶⊕i=1N\CalAi,
where \CalAi=\CalA,i=1,…,N
defined by
ξρ(a)=⊕i=1Nραi(a)
is injective.
Consider the homomorphisms:
[TABLE]
Since
\CalB⊗ξρ(\CalA)
is a subalgebra of
\CalB⊗(⊕i=1N\CalAi)=⊕i=1N(\CalB⊗\CalAi)
and
both idB⊗ξρ
and ⊕i=1N(αi⊗id) are isomorphisms,
the composition
⊕i=1N(αi⊗id)∘(id⊗ξρ)
is isomorphic.
Hence
[TABLE]
is injective.
This implies that
{ραiΣ⊗}i=1N is faithful.
By the equality
[TABLE]
for αi1,⋯,αin∈Σ,
the presented subshifts
ΛρΣ⊗
and
Λρ
coincide.
∎
We denote by
Λ the presented subshift
Λρ(=ΛρΣ⊗).
Let Sαi be the generating partial isometries of \CalA⋊ρΛ
satisfying
Sαi∗xSαi=ραi(x)
for
x∈\CalA,i=1,…,N,
and
Sαi those of
(\CalB⊗\CalA)⋊ρΣ⊗Λ
satisfying
Sαi∗ySαi=ραiΣ⊗(y)
for
y∈\CalB⊗\CalA,i=1,…,N.
Proposition 6.2
There exists a unital embedding ι~
of \CalA⋊ρΛ into (\CalB⊗\CalA)⋊ρΣ⊗Λ in a canonical way.
Demonstration Proof
Define the injective homomorphism
ι:\CalA→\CalB⊗\CalA by setting
ι(a)=1⊗a for a∈A.
Since the equality
Sαi∗ι(a)Sαi=ι(ραi(a))
for
a∈\CalA,i=1,…,N
holds,
there exists a homomorphism
ι~ from \CalA⋊ρΛ to (\CalB⊗\CalA)⋊ρΣ⊗Λ
satisfying
ι~(a)=1⊗a,ι~(Sαi)=Sαi
for
a∈\CalA,i=1,…,N
by the universality of \CalA⋊ρΛ.
Let
\CalEρ:\CalA⋊ρΛ→\CalFρ
and
\CalEρΣ⊗:(\CalB⊗\CalA)⋊ρΣ⊗Λ→\CalFρΣ⊗
be the canonical conditional expectations respectively.
Define the C∗-subalgebras
\CalF(C⊗\CalA,ρΣ⊗)⊂(C⊗\CalA)⋊ρΣ⊗Λ
of
(\CalB⊗\CalA)⋊ρΣ⊗Λ
by setting
[TABLE]
The diagrams
[TABLE]
are commutative.
Since ι:\CalA→C⊗\CalA is isomorphic,
so is
the restriction
ι~∣\CalFρ:\CalFρ→\CalF(C⊗\CalA,ρΣ⊗)
of
\CalFρ.
One indeed sees that
the condition
SμaSν∗=0
for some
a∈\CalA,∣μ∣=∣ν∣
implies
Sμ(1⊗a)Sν∗=0
because of the equality
ι(ρμ(1)aρν(1))=Sμ∗Sμ(1⊗a)Sν∗Sν.
For
∑μ,ν∈ΛkSμaμ,νSν∗∈\CalFρ,
suppose that
ι~(∑μ,ν∈ΛkSμaμ,νSν∗)=0.
It follows that for any ξ,η∈Λk,
[TABLE]
so that
0=Sξ(1⊗aξ,η)Sη∗,
and hence
Sξaξ,ηSν∗=0.
This implies that
ι~∣\CalFρk:\CalFρk→\CalF(C⊗\CalA,ρΣ⊗)k
is injective
and so is
ι~∣\CalFρ:\CalFρ→\CalF(C⊗\CalA,ρΣ⊗).
Therefore by using a routine argument,
one concludes that
ι~:\CalA⋊ρΛ→(C⊗\CalA)⋊ρΣ⊗Λ
is injective and hence isomorphic.
∎
Let us prove that (\CalB⊗\CalA,ρΣ⊗,Σ)
satisfies condition (I)
if
(\CalA,ρ,Σ)
satisfies condition (I).
The result will be used in the following sections.
We set the C∗-subalgebras
\CalD(C⊗\CalA,ρΣ⊗)⊂\CalDρΣ⊗
of
\CalFρΣ⊗
by setting
[TABLE]
We may identify
the subalgebra \CalDρ of \CalFρ
with the subalgebra
\CalD(C⊗\CalA,ρΣ⊗)
of
\CalF(C⊗\CalA,ρΣ⊗)
through the map ι~ as in the preceding proposition.
Let φ∈\CalB∗ be a faithful state on \CalB.
It is well-known that there exists a faithful projection
Θφ:\CalB⊗\CalA→\CalA of norm one satisfying
Θφ(b⊗a)=φ(b)a for b⊗a∈\CalB⊗\CalA.
Lemma 6.3
Let φ∈\CalB∗ be a faithful state on \CalB
satisfying φ∘αi=φ,i=1,…,N.
The projection Θφ:\CalB⊗\CalA→\CalA
of norm one can be extended to a projection of norm one
Θ\CalD:\CalDρΣ⊗→\CalDρ
such that
Θ\CalD(x)=x for x∈\CalDρ.
Demonstration Proof
For k∈N, define the C∗-subalgebras
\CalDρk of \CalDρ
and
\CalDρΣ⊗k
of
\CalDρΣ⊗
by setting
[TABLE]
For xμ∈\CalB⊗\CalA,ξ∈Λk,
the identities
[TABLE]
hold,
so that the map defined by
Θ\CalDk:\CalDρΣ⊗k→\CalDρk
[TABLE]
is well-defined for each k∈Z+.
We will next see the restriction of
Θ\CalDk+1 to \CalDρΣ⊗k coincides with
Θ\CalDk.
Since
∑μ∈ΛkSμxμSμ∗∈\CalDρΣ⊗k
is written as
∑μ∈Λk∑i=1NSμSαiSαi∗xμSαiSαi∗Sμ∗∈\CalDρΣ⊗k+1,
it follows that
[TABLE]
As the state φ is αi-invariant
for
i=1,…,N,
one has
for ∑jbj⊗aj∈\CalB⊗\CalA,
[TABLE]
so that
Θφ(ραiΣ⊗(xμ))=Sαi∗Θφ(xμ)Sαi
for xμ∈\CalB⊗\CalA.
It then follows that
[TABLE]
Therefore the sequence
{Θ\CalDk}k=1∞
defines a projection from
\CalDρΣ⊗ onto \CalDρ, which we denote by Θ\CalD.
∎
Lemma 6.4
Assume that (\CalA,ρ,Σ) is central.
Then
Sμ(1⊗a)Sμ∗
commutes with b⊗1 for
a∈\CalA,μ∈Λ∗ and b∈\CalB.
Demonstration Proof
Since
(1⊗a)ρμΣ⊗(b⊗1)=ρμΣ⊗(b⊗1)(1⊗a),
it follows that
[TABLE]
∎
Theorem 6.5
Assume that there exists a faithful state φ on \CalB
invariant under αi∈Aut(\CalB),i=1,…,N.
Suppose that (\CalA,ρ,Σ) is central.
If (\CalA,ρ,Σ) satisfies condition (I),
then
(\CalB⊗\CalA,ρΣ⊗,Σ)
satisfies condition (I) and is central.
Demonstration Proof
Since (\CalA,ρ,Σ) satisfies condition (I),
there exists a increasing sequence
\CalAl,l∈Z+
of C∗-subalgebras of \CalA and
a projection
qkl∈\CalDρ∩\CalAl′
with l≥k
satisfying the conditions of condition (I).
We set
(\CalB⊗\CalA)l=\CalB⊗\CalAl,l∈Z+.
Then the conditions
∪l∈N(\CalB⊗\CalA)l=\CalB⊗\CalA
and
ραiΣ⊗((\CalB⊗\CalA)l)⊂(\CalB⊗\CalA)l+1
are easy to veryfy.
Let ι~:\CalA⋊ρΛ↪(\CalB⊗\CalA)⋊ρΣ⊗Λ
be the embedding in Proposition 6.2.
Put
q~kl=ι~(qkl)∈\CalDρΣ⊗
for l≥k.
By the preceding lemma, one sees that
q~kl∈\CalDρΣ⊗∩((\CalB⊗\CalA)l)′.
We will show that
q~klx=0 for 0=x∈(\CalB⊗\CalA)l.
As xx∗∈\CalB⊗\CalAl, one has
Θ\CalD(xx∗)=Θφ(xx∗)∈\CalAl.
Hence
qklΘφ(xx∗)=0.
By the equality
Θ\CalD(q~klxx∗q~kl)=qklΘφ(xx∗)qkl,
one obtains
q~klx=0.
Let
ϕ~ρΣ⊗(X)=∑i=1NSαiXSαi∗
for
X∈\CalDρΣ⊗.
One has
[TABLE]
Thus
(\CalB⊗\CalA,ρΣ⊗,Σ)
satisfies condition (I).
If (\CalA,ρ,Σ) is central,
the projections 1⊗ρμ(1) for μ∈Λ∗
commute with \CalB⊗\CalA,
so that
(\CalB⊗\CalA,ρΣ⊗,Σ)
is central.
∎
We will study structure of the fixed point algebra
\CalFρΣ⊗
of
(\CalB⊗\CalA)⋊ρΣ⊗Λ
under the gauge action
ρΣ⊗.
Recall that \CalFρ denote the fixed point algebra of \CalA⋊ρΛ
under the gauge action