1: \documentclass{elsart}
2:
3: \usepackage{amssymb}
4: \usepackage{amsmath}
5:
6:
7: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
8: % Private definitions
9:
10: %abbreviation "l.c." for locally compact
11: \newcommand{\lc}{l.c.}
12: %abbreviation "n.s.f." for normal semifnite faithful
13: \newcommand{\nsf}{n.s.f.}
14: %quotient of two groups
15: \newcommand{\quotient}[2]{\raisebox{-0.195cm}[0.38cm]{}\mbox{\raisebox{-1pt}
16: {$#1/\hspace{-2pt}\raisebox{-0.05cm}{${\textstyle#2}$}
17: $}}}
18: %caracteristic function of a set
19: \def\11{\mbox1\hspace{-.25em}\text{I}}
20: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
21:
22: \journal{Journal of Functional Analysis}
23:
24: \begin{document}
25: \begin{frontmatter}
26: \title{On locally compact quantum groups\\
27: whose algebras are factors}
28: \author{Pierre Fima}
29: \address{Laboratoire de math\'{e}matiques Nicolas Oresme, Universit\'{e} de Caen BP 5186, F14032 Caen Cedex, France}
30: \ead{pierre.fima@math.unicaen.fr}
31:
32: \begin{abstract}
33: In this paper we are interested in examples of locally compact quantum groups $(M,\Delta)$ such that both von Neumann algebras, $M$ and the dual $\hat{M}$, are factors. There is a lot of known examples such that $(M,\hat{M})$ are respectively of type $(\rm{I}_{\infty},\rm{I}_{\infty})$ but there is no examples with factors of other types. We construct new examples of type $(\rm{I}_{\infty},\rm{II}_{\infty})$, $(\rm{II}_{\infty},\rm{II}_{\infty})$ and $(\rm{III}_{\lambda},\rm{III}_{\lambda})$ for each $\lambda\in [0,1]$. Also we show that there is no such example with $M$ or $\hat{M}$ a finite factor.
34: \end{abstract}
35: \begin{keyword}
36: Quantum groups \sep Factors \sep Crossed product
37: \MSC 46L52 \sep 46L65
38: \end{keyword}
39:
40: \end{frontmatter}
41:
42: \section{Introduction}
43: A locally compact (\lc) quantum group, in the von Neumann algebra setting (see \cite{KusVa2,KusVa3}), is a pair $(M,\Delta)$, where $M$ is a von Neumann algebra and $\Delta$ is a comultiplication on $M$, with left and right invariant weight. In a canonical way, every \lc{} group is a commutative \lc{} quantum group and every von Neumann group algebra of a \lc{} group is a cocommutative \lc{} quantum group. Conversely, every commutative or cocommutative \lc{} quantum group is obtained in this way. Our aim is to obtain new examples of \lc{} quantum groups which are as far as possible from groups so we will be interested in the "least commutative and cocommutative" examples. The formulation of the problem is the following. Given a pair of factors of a certain type $(x_{1}, x_{2})$, is it possible to find a \lc{} quantum group $(M,\Delta)$ such that $M$ is a type $x_{1}$ factor and the dual algebra $\hat{M}$ is a type $x_{2}$ factor (in the sense of Murray-von Neumann's and Connes
44: ' classification of factors) ? There exists a lot of examples for the case $(\rm{I}_{\infty},\rm{I}_{\infty})$. In particular, it is shown in \cite{BaaSkVa} that for any matched pair of conjugated \lc{} groups the bicrossed product is a type $(\rm{I}_{\infty},\rm{I}_{\infty})$ \lc{} quantum group.
45:
46: We start with a negative result showing that, if $(M,\Delta)$ is a \lc{} quantum group and $M$ is a finite factor, then $\hat{M}$ is not a factor. On a positive side, our tool to construct examples of \lc{} quantum groups is to use the bicrossed product of \lc{} groups (see \cite{VaesVain}). One can show that, if an action of a \lc{} group on a von Neumann algebra is free on its center, then the type of the cocycle crossed product does not depend a lot on the cocycle. This is why we only consider trivial cocycles. We construct for any $\lambda\in [0,1]$ a \lc{} quantum group $(M,\Delta)$ such that $M$ and $\hat{M}$ are type $\rm{III}_{\lambda}$ factors. Similary we obtain examples of type $(\rm{I}_{\infty},\rm{II}_{\infty})$ and $(\rm{II}_{\infty},\rm{II}_{\infty})$. All of them are ITPFI factors, a kind of infinite tensor product of the p-adic version of the Baaj and Skandalis' example (see \cite{VaesVain}).
47:
48: This paper is organized as follows. In the second section we introduce some notations and recall some elementary facts about \lc{} quantum groups, bicrossed product construction, infinite tensor product and ITPFI factors. In the third section we prove that a \lc{} quantum group $(M,\Delta)$ with $M$ a finite factor is compact. In the fourth section we describe our examples using $p$-adic numbers.
49:
50: \section{Preliminaries}
51: \paragraph{Locally compact quantum groups}
52:
53: In this paper we suppose that all von Neumann algebras have separable predual and \lc{} groups are second countable. We denote by $\otimes$ the tensor product of Hilbert spaces or von Neumann algebras. We refer to \cite{Str2} for the theory of normal semifinite faithful (\nsf) weights on von Neumann algebras. If $\varphi$ is a \nsf{} weight on $M$, we use the standard notation
54: \begin{displaymath}
55: \mathcal{M}_{\varphi}^{+}=\{x\in M^{+}\,|\,\varphi(x)<\infty\},\,\,
56: \mathcal{N}_{\varphi}=\{x\in M\,|\,x^{*}x\in\mathcal{M}_{\varphi}^{+}\},
57: \,\, \mathcal{M}_{\varphi}=\mathcal{N}_{\varphi}^{*}\mathcal{N}_{\varphi}.
58: \end{displaymath}
59: We use \lc{} quantum groups in the von Neumann algebraic setting (see \cite{KusVa2}). A pair $(M,\Delta)$ is called a \lc{} quantum group when
60: \begin{itemize}
61: \item $M$ is a von Neumann algebra and $\Delta\rightarrow M\otimes M$ is a normal and unital *-homomorphism satisfying the coassociativity condition :
62: \begin{displaymath}
63: (\Delta\otimes\iota)\Delta=(\iota\otimes\Delta)\Delta,
64: \end{displaymath}
65: where $\iota$ is the identity map.
66: \item There exist \nsf{} weights $\varphi$ and $\psi$ on $M$ such that
67: \begin{itemize}
68: \item $\varphi$ is left invariant in the sense that
69: \begin{displaymath}
70: \varphi\left((\omega\otimes\iota)\Delta(x)\right)=\varphi(x)\omega(1),\,\,\forall x\in\mathcal{M}_{\varphi}^{+},\,\,\forall \omega\in M_{*}^{+},
71: \end{displaymath}
72: \item $\psi$ is right invariant in the sense that
73: \begin{displaymath}
74: \psi\left((\iota\otimes\omega)\Delta(x)\right)=\psi(x)\omega(1),\,\,\forall x\in\mathcal{M}_{\psi}^{+},\,\,\forall\omega\in M_{*}^{+}.
75: \end{displaymath}
76: \end{itemize}
77: \end{itemize}
78: From \cite{KusVa2} we know that left invariant weights on $(M,\Delta)$ are unique to a positive scalar and the same holds for right invariant weights.
79:
80: A \lc{} quantum group is called compact if its left invariant weight is finite.
81:
82: Let $(M,\Delta)$ be a \lc{} quantum group, fix a left invariant \nsf{} weight $\varphi$ on $(M,\Delta)$ and represent $M$ on the GNS-space of $\varphi$ such that $(H,\iota,\Lambda)$ is a GNS construction for $\varphi$. Then we can define a unitary $W$ on $H\otimes H$ by
83: \begin{displaymath}
84: W^{*}\left(\Lambda(a)\otimes\Lambda(b)\right)=(\Lambda\otimes\Lambda)\left(\Delta(b)(a\otimes 1)\right)\qquad\text{for all }a,b\in\mathcal{N}_{\varphi}.
85: \end{displaymath}
86: Where $\Lambda\otimes\Lambda$ is the canonical GNS-map for the tensor product weight $\varphi\otimes\varphi$. $W$ is called the fondamental unitary of $(M,\Delta)$. The comultiplication can be given in terms of $W$ by the formula $\Delta(x)=W^{*}(1\otimes x)W$ for all $x\in M$. Also the von Neumann algebra $M$ can be written in terms of $W$ as
87: \begin{displaymath}
88: M=\{(\iota\otimes\omega)(W)|\omega\in\mathcal{B}(H)_{*}\}^{-\sigma\mathrm{-strong}*},
89: \end{displaymath}
90: where $\{X\}^{-\sigma\mathrm{-strong}*}$ denote the $\sigma$-strong* closure of $X$. It is possible to define a new von Neumann algebra
91: \begin{displaymath}
92: \hat{M}=\{(\omega\otimes\iota)(W)|\omega\in\mathcal{B}(H)_{*}\}^{-\sigma\mathrm{-strong}*},
93: \end{displaymath}
94: and a comultiplication on $\hat{M}$ by $\hat{\Delta}(x)=\Sigma W(x\otimes 1)W^{*}\Sigma$ for $x\in\hat{M}$, where $\Sigma$ is the flip map on $H\otimes H$. Also, one can construct left and right invariant weight on $\hat{M}$ for $\hat{\Delta}$. We obtain in this way a new \lc{} quantum group $(\hat{M},\hat{\Delta})$ called the dual of $(M,\Delta)$. From \cite{KusVa2} we know that the bidual quantum group $(\hat{\hat{M}},\hat{\hat{\Delta}})$ is isomorphic to $(M,\Delta)$.
95:
96: If $G$ is an ordinary locally compact group then $M=L^{\infty}(G)$ with the comultiplication $\Delta_{G}(f)(s,t)=f(st)$ and left and right invariant weight obtained by left and right Haar measure is a commutative \lc{} quantum group. Every commutative \lc{} quantum group is obtained in this way. Now, take $\hat{M}=\mathcal{L}(G)$ the group von Neumann algebra and $\hat{\Delta}_{G}(\lambda_{g})=\lambda_{g}\otimes\lambda_{g}$ where $(\lambda_{g})$ is the left regular representation of $G$. One can show that the Plancherel weight on $\hat{M}$ is left invariant. Also, it is easy to obtain a right invariant weight on $\hat{M}$ so $(\mathcal{L}(G),\hat{\Delta}_{G})$ is a \lc{} quantum group, this the dual of $(L^{\infty}(G),\Delta_{G})$. It is obvious that $\sigma\hat{\Delta}_{G}=\hat{\Delta}_{G}$ where $\sigma$ is the flip map on $M\otimes M$. We say that $(\mathcal{L}(G),\hat{\Delta}_{G})$ is cocommutative. One can show that every cocommutative \lc{} quantum group is obtained in this way.
97:
98: \paragraph{Bicrossed product}
99:
100: Let $G_{1}, G_{2}$ be two closed subgroups of a \lc{} group $G$ such that $G_{1}\cap G_{2}=\{e\}$ and $\mu(G-G_{1}G_{2})=0$ where $e$ is the identity element of $G$ and $\mu$ is a Haar measure on $G$. We say that the pair $(G_{1}, G_{2})$ is matched. We will now describe the bicrossed product construction of \lc{} group (see \cite{VaesVain}). Let $(G_{1}, G_{2})$ be a matched pair of \lc{} groups and $g\in G_{1}$, $s\in G_{2}$, then we can write nearly everywhere
101: \begin{displaymath}
102: gs=\alpha_{g}(s)\beta_{s}(g).
103: \end{displaymath}
104: We obtain two maps defined nearly everywhere and measurable
105: \begin{align*}
106: \alpha\,:\,G_{1}\times G_{2}\rightarrow G_{2}\,:\qquad (g,s)\rightarrow\alpha_{g}(s), \\
107: \beta\,:\,G_{2}\times G_{1}\rightarrow G_{1}\,:\qquad (s,g)\rightarrow\beta_{s}(g).
108: \end{align*}
109: Now we define two normal unital *-homomorphisms
110: \begin{align*}
111: \alpha\, : \, L^{\infty}(G_{2})\rightarrow L^{\infty}(G_{1}\times G_{2})\,\, : \,\, (\alpha f)(g,s)=f(\alpha_{g}(s)),\\
112: \beta\, : \, L^{\infty}(G_{1})\rightarrow L^{\infty}(G_{2}\times G_{1})\,\, : \,\, (\beta f)(s,g)=f(\beta_{s}(g)),
113: \end{align*}
114: and we have
115: \begin{displaymath}
116: (\iota\otimes\alpha)\alpha=(\Delta_{G_{1}}\otimes\iota)\alpha\qquad\text{and}\qquad
117: (\iota\otimes\beta)\beta=(\Delta_{G_{2}}\otimes\iota)\beta.
118: \end{displaymath}
119: Hence $\alpha$ will be an action of $G_{1}$ on the von Neumann algebra $L^{\infty}(G_{2})$ and $\beta$ an action of $G_{2}$ on the von Neumann algebra $L^{\infty}(G_{1})$. So we can define the crossed product von Neumann algebra $M=G_{1}\ltimes L^{\infty}(G_{2})$ and a faithfull *-homomorphism $\Delta$ : $M\rightarrow M\otimes M$
120: \begin{displaymath}
121: \Delta(x)=W^{*}(1_{L^{2}(G_{1}\times G_{2})}\otimes x)W\qquad\text{for }x\in M,
122: \end{displaymath}
123: where $W$ is a unitary in $L^{2}(G_{1}\times G_{2}\times G_{1}\times G_{2})$ defined by
124: \begin{displaymath}
125: (W\xi)(g,s,h,t)=\xi(\beta_{\alpha_{g}(s)^{-1}t}(h)g,s,h,\alpha_{g}(s)^{-1}t).
126: \end{displaymath}
127: Then one can prove that $\Delta$ is a comultiplication on $M$ and the dual weight of the left invariant integral on $L^{\infty}(G_{2})$ is left invariant for $\Delta$. In this way we obtain a \lc{} quantum group $(M,\Delta)$ with dual $(\hat{M},\hat{\Delta})$ such that
128: \begin{displaymath}
129: \hat{M}=L^{\infty}(G_{1})\rtimes G_{2},\quad
130: \hat{\Delta}(x)=\Sigma W(x\otimes 1_{L^{2}(G_{1}\times G_{2})})W^{*}\Sigma\quad\text{for }x\in \hat{M}.
131: \end{displaymath}
132:
133: \paragraph{Infinite tensor product of von Neumann algebras}
134:
135: For each $n\in\Nset$ let $M_{n}$ be a von Neumann algebra acting on an Hilbert space $H_{n}$ and $\xi_{n}$ a norm $1$ vector in $H_{n}$. The infinite tensor product of $M_{n}$ relatively to $\xi_{n}$ is the von Neumann algebra generated by the operators $x_{1}\otimes\ldots\otimes x_{k}\otimes 1\otimes\ldots$ for $k\in \Nset$ and $x_{i}\in
136: M_{i}$ in the infinite tensor product of Hilbert spaces $H_{n}$ relatively to the norm $1$ vectors $\xi_{n}$. We denote this von Neumann algebra by
137: \begin{displaymath}
138: \bigotimes(M_{n},H_{n},\xi_{n}).
139: \end{displaymath}
140: If each $M_{n}$ is a factor then $\bigotimes(M_{n},H_{n},\xi_{n})$ is a factor (see \cite{ArW}).
141:
142: The following lemma is certainly well known but we could not find a proof in the literature.
143: \begin{lem}\label{valprop}
144: Let $p_{n}\neq 0$ be a projection in a von Neumann algebra $M_{n}$ and $\omega_{n}$ a normal faithful state on $M_{n}$ with GNS space $H_{n}$ such that $\omega_{n}=\omega_{\xi_{n}}$. Put $M=\bigotimes\left(M_{n},\omega_{n}\right)$. The decreasing sequence of projections
145: \begin{displaymath}
146: p_{1}\otimes\ldots\otimes p_{n}\otimes 1\ldots
147: \end{displaymath}
148: converge to a projection $p\in M$ and we have
149: \begin{displaymath}
150: p\neq 0\Leftrightarrow\sum_{n}||(1-p_{n})\xi_{n}||^{2}<\infty,
151: \end{displaymath}
152: and if $p\neq 0$
153: \begin{displaymath}
154: pMp\simeq
155: \bigotimes\left(p_{n}M_{n}p_{n},\omega_{\eta_{n}}\right),\quad\text{where}\quad \eta_{n}:=\frac{p_{n}\xi_{n}}{||p_{n}\xi_{n}||}.
156: \end{displaymath}
157: \end{lem}
158: \begin{pf}
159: $p\neq 0$ if and only if $(\otimes\omega_{n})(p)>0$. This is equivalent to
160: \begin{displaymath}
161: \sum_{n}||(1-p_{n})\xi_{n}||^{2}<\infty.
162: \end{displaymath}
163: The isomorphism is a simple identification.\qed
164: \end{pf}
165:
166: \begin{exmp}
167: For each $n\in\Nset$ let $M_{n}$ be a von Neumann algebra, $\varphi_{n}$ a \nsf{} weight on $M_{n}$ and $q_{n}\in M_{n}$ a projector with $\varphi_{n}(q_{n})=1$. We can take $H_{n}$ the Hilbert space of the G.N.S. construction for $\varphi_{n}$ and $\xi_{n}=\Lambda_{\varphi_{n}}(q_{n})$. We introduce the notation
168: \begin{displaymath}
169: M=\bigotimes\left(M_{n},\varphi_{n},q_{n}\right):= \bigotimes(M_{n},H_{n},\xi_{n}).
170: \end{displaymath}
171: Observe that there is a natural projector $q\in M$ defined by the infinite tensor product of the $q_{n}$. When $\varphi_{n}=\omega_{n}$ is a normal faithful state on $M_{n}$ and $q_{n}=1$, we use the standard notation
172: \begin{displaymath}
173: \bigotimes\left(M_{n},\omega_{n}\right):=\bigotimes\left(M_{n},\omega_{n},1\right).
174: \end{displaymath}
175: \end{exmp}
176:
177: \begin{prop}\label{qMq}
178: Let $M_{n}$, $\varphi_{n}$ and $q_{n}$ be as above and suppose that $q_{n}\in M_{n}^{\varphi_{n}}$. Denote by $\omega_{\xi_{n}}$ the vector state associated to $\xi_{n}$. Then $\omega_{\xi_{n}}$ is faithful on $q_{n}M_{n}q_{n}$ and
179: \begin{displaymath}
180: qMq\simeq\bigotimes\left(q_{n}M_{n}q_{n},\omega_{\xi_{n}}\right).
181: \end{displaymath}
182: \end{prop}
183:
184: \begin{pf}
185: Let $J_{n}$ be the usual antiunitary operator associated with $\varphi_{n}$ and $J=\otimes J_{n}$, $q=\otimes q_{n}$. We have $||x\xi_{n}||^{2}=\varphi_{n}(q_{n}x^{*}xq_{n})$ thus the faithfulness of $\varphi_{n}$ on $M_{n}$ implies that $\omega_{\xi_{n}}$ is faithful on $q_{n}M_{n}q_{n}$. Note that the close linear subspace of $q_{n}H_{n}$ generated by $q_{n}M_{n}q_{n}\xi_{n}$ is $J_{n}q_{n}J_{n}q_{n}H_{n}$. Thus, the GNS space of $\omega:=\otimes\omega_{\xi_{n}}$ on $N:=\bigotimes\left(q_{n}M_{n}q_{n},\omega_{\xi_{n}}\right)$ is canonically isomorphic with $qJqJH$, and the image of $qMq$ by the restriction homomorphism to the invariant subspace $qJqJ H$ is $N$. This homomorphism is in fact an isomorphism because the closure of $M^{'}qJqJH$ is $qH$. \qed
186: \end{pf}
187:
188: \begin{rem}
189: One can show that if $q_{n}\in M_{n}^{\varphi_{n}}$ there exists a canonical \nsf{} weight $\varphi$ on $M$ such that $\sigma_{t}^{\varphi}=\otimes_{n}\sigma_{t}^{\varphi_{n}}$ and $\varphi q =\otimes_{n} \varphi_{n} q_{n}$. This is the noncommutative analogue of the restricted direct product of measurable spaces with non necessarily finite measure.
190: \end{rem}
191:
192: \paragraph{Resctricted direct product action}
193:
194: In the sequel, all group actions on von Neumann algebras are
195: supposed to be pointwise $\sigma$-weakly continuous. Let $G_{n}$ be
196: a sequence of \lc{} groups and $\mu_{n}$ a left Haar measure on
197: $G_{n}$. We suppose that for all $n$ there is a compact open
198: subgroup $K_{n}$ of $G_{n}$ such that $\mu_{n}(K_{n})=1$. Recall
199: that the restricted direct product $\prod\,^{'}(G_{n},K_{n})$ is
200: defined as the set of $(x_{n})\in\prod G_{n}$ such that $x_{n}\in
201: K_{n}$ for $n$ large enough (see \cite{Bl} for details). Let $M_{n}$
202: be a sequence of von Neumann algebras with actions
203: $\alpha^{n}\,:\,G_{n}\rightarrow \text{Aut}(M_{n})$. Let
204: $\varphi_{n}$ be a \nsf{} weight on $M_{n}$ and $q_{n}$ a projection
205: in the centralizer of $\varphi_{n}$ with $\varphi_{n}(q_{n})=1$ and such that for all $n$, for all
206: $g_{n}\in G_{n}$, there exists $c_{n}(g_{n})>0$ such that
207: $\varphi_{n}\circ\alpha^{n}=c_{n}(g_{n})\varphi_{n}$,
208: $c_{n}|_{K_{n}}=1$ and for all $g_{n}\in K_{n}$ one has
209: $\alpha^{n}_{g_{n}}(q_{n})=q_{n}$. With this data one can construct
210: an action of the restricted direct product
211: $G=\prod^{'}\left(G_{n},K_{n}\right)$ on the infinite tensor product $M=\bigotimes\left(M_{n},\varphi_{n},q_{n}\right)$. We fix a G.N.S. construction $(H_{n},\iota,\Lambda_{n})$ for $\varphi_{n}$ and we put $H=\bigotimes\left(H_{n},\Lambda_{n}(q_{n})\right)$.
212:
213: \begin{prop}
214: There exists a unique action $\alpha\,:\,G\rightarrow\text{Aut}(M)$, called the restricted direct product of $\alpha^{n}$, such that for all $g=(g_{n})\in G$ and $x_{n}\in M_{n}$
215: \begin{displaymath}
216: \alpha_{g}(x_{1}\otimes\ldots\otimes x_{n}\otimes 1\ldots)
217: =\alpha^{1}_{g_{1}}(x_{1})\otimes\ldots\otimes\alpha^{n}_{g_{n}}(x_{n})\otimes 1\otimes\ldots
218: \end{displaymath}
219: \end{prop}
220:
221: \begin{pf}
222: The uniqueness part is obvious. To show the existence, we first compute a unitary implementation of the actions $\alpha^{n}$. It is easy to see that, for $g_{n}\in G_{n}$, the operator
223: \begin{displaymath}
224: U_{n}(g_{n})\,:\,\Lambda_{n}(x)\mapsto c_{n}(g_{n})^{\frac{1}{2}}\Lambda_{n}(\alpha^{n}_{g_{n}}(x)),\qquad\text{for }x\in\mathcal{N}_{\varphi_{n}},
225: \end{displaymath}
226: can be extended to a unitary operator on $H_{n}$, still denoted by $U_{n}(g_{n})$, and such that
227: \begin{displaymath}
228: \alpha^{n}_{g_{n}}(x)=U_{n}(g_{n})xU_{n}(g_{n})^{*}\qquad\text{for all }x\in M_{n}.
229: \end{displaymath}
230: The hypotesis implies that for all $n$ and for all $g_{n}\in K_{n}$ one has $U_{n}(g_{n})\Lambda_{n}(q_{n})=\Lambda_{n}(q_{n})$. Then for all $g\in G$ one can define a unitary operator $U_{g}$ on $H$ by $U_{g}=\otimes_{n} U_{n}(g_{n})$ where $g=(g_{n})$. In this way we obtain a group homomorphism $g\mapsto U_{g}$ from $G$ to the unitary group of $H$. Because we have $U_{g} M U_{g}^{*}=M$ this allows us to construct a group homomorphism $\alpha\,:G\rightarrow\text{Aut}(M)$ defined by
231: \begin{displaymath}
232: \alpha_{g}(x)=U_{g} x U_{g}^{*}\qquad\text{for }x\in M.
233: \end{displaymath}
234: This is obvious that $\alpha$ is pointwise $\sigma$-weakly continuous and verifies the equation.\qed
235: \end{pf}
236:
237: Let us identify the crossed product of $G$ by $M$ with an infinite tensor product of the crossed products of $G_{n}$ by $M_{n}$. We denote by $\pi_{n}$ the inclusion of $M_{n}$ into $G_{n}\ltimes M_{n}$ and $\pi$ the inclusion of $M$ into $G\ltimes M$. We denote by $\tilde{\varphi}_{n}$ the dual weight of $\varphi_{n}$ and by $\11_{A}$ the caracteristic function of a mesurable set $A$.
238:
239: \begin{prop}\label{crossed}
240: Let $e_{n}=\pi_{n}(q_{n})\left(\lambda(\11_{K_{n}})\otimes 1\right)$, where $\lambda(\11_{K_{n}})$ is the convolution operator by $\11_{K_{n}}$. Then $e_{n}$ is a projection in $G_{n}\ltimes M_{n}$. Moreover, one has $\tilde{\varphi}_{n}(e_{n})=1$, $e_{n}\in \left(G_{n}\ltimes M_{n}\right)^{\tilde{\varphi}_{n}}$ and
241: \begin{displaymath}
242: G\ltimes M\simeq
243: \bigotimes\left(G_{n}\ltimes M_{n},\tilde{\varphi}_{n},e_{n}\right).
244: \end{displaymath}
245: \end{prop}
246:
247: \begin{pf}
248: Because $\pi_{n}(q_{n})$ and $\lambda(\11_{K_{n}})$ are projections, if $\pi_{n}(q_{n})$ and $\lambda(\11_{K_{n}})\otimes 1$ commute then $e_{n}$ is a projection. Take $\xi\in L^{2}(G_{n},H_{n})$ then
249: \begin{eqnarray*}
250: \lefteqn{\left((\lambda(\11_{K_{n}})\otimes 1)\pi_{n}(q_{n})\xi\right)(g)=\int_{G_{n}}\11_{K_{n}}(t)\alpha^{n}_{g^{-1}t}(q_{n})\xi(t^{-1}g)\d\mu_{n}(t)}\\
251: & & =\alpha^{n}_{g^{-1}}(q_{n})\int_{G_{n}}\11_{K_{n}}(t)\xi(t^{-1}g)\d\mu_{n}(t),\,\,\,\text{because}\,\,\,\forall t\in K_{n},\,\,\,\alpha^{n}_{t}(q_{n})=q_{n}\\
252: & & =\left(\pi_{n}(q_{n})(\lambda(\11_{K_{n}})\otimes1)\xi\right)(g).
253: \end{eqnarray*}
254: Thus $e_{n}$ is a projection. Now, using $K_{n}\subset\text{Ker}(\delta_{G_{n}})$ and $c_{n}|_{K_{n}}=1$, we have
255: \begin{displaymath}
256: \sigma_{t}^{\tilde{\varphi}_{n}}(\lambda(\11_{K_{n}})\otimes1)=\lambda(\11_{K_{n}})\otimes1.
257: \end{displaymath}
258: This implies that
259: \begin{displaymath}
260: \sigma_{t}^{\tilde{\varphi}_{n}}(e_{n})=\pi_{n}(\sigma_{t}^{\varphi_{n}}(q_{n}))
261: \sigma_{t}^{\tilde{\varphi}_{n}}(\lambda(\11_{K_{n}})\otimes1)=e_{n}.
262: \end{displaymath}
263: Next, using definition of the dual weight, we have $\tilde{\varphi}_{n}(e_{n})=\varphi_{n}(q_{n})\11_{K_{n}}(1)
264: =1$. Recall that, using the classical explicit G.N.S. construction $(L^{2}(G_{n},M_{n}),\iota,\tilde{\Lambda}_{n})$ for the dual weight, one has (see \cite{Vaes})
265: \begin{displaymath}
266: \tilde{\Lambda}_{n}(e_{n})=\11_{K_{n}}\otimes\Lambda_{n}(q_{n}).
267: \end{displaymath}
268: We denote this vector by $\xi_{n}$. We define the operator
269: \begin{displaymath}
270: U\,:\,\bigotimes\left(L^{2}(G_{n},H_{n}),\xi_{n}\right)
271: \rightarrow L^{2}(G,H)
272: \end{displaymath}
273: on a dense subset by
274: \begin{displaymath}
275: U(F_{1}\otimes\ldots\otimes F_{n}\otimes\bar{\xi}_{n})(g)
276: =\bigotimes_{i=1}^{n}F_{i}(g_{i})\otimes\left(
277: \bigotimes_{i=n+1}^{\infty}\11_{K_{i}}(g_{i})\Lambda_{i}(p_{i})\right),
278: \end{displaymath}
279: where $F_{i}\in L^{2}(G_{i},H_{i})$, $g=(g_{n})\in G$ and $\bar{\xi}_{n}=\otimes_{i=n+1}^{\infty}\xi_{i}$. Then $U$ is an isometry with dense range. Thus we obtain a unitary operator, again denoted by $U$, such that, if $g=(g_{1},\ldots,g_{n},1,\ldots)$,
280: \begin{displaymath}
281: U\left((\lambda_{g_{1}}\otimes1)\otimes\ldots\otimes(\lambda_{g_{n}}\otimes 1)\otimes 1\otimes\ldots\right) U^{*}
282: =\lambda_{g}\otimes 1,
283: \end{displaymath}
284: \begin{displaymath}
285: U\left(\pi_{1}(x_{1})\otimes\ldots\otimes\pi_{n}(x_{n})\otimes 1\otimes\ldots\right) U^{*}
286: =\pi(x_{1}\otimes\ldots\otimes x_{n}\otimes 1\otimes\ldots).
287: \end{displaymath}
288: It follows that
289: \begin{displaymath}
290: U\left(\bigotimes\left(G_{n}\ltimes M_{n},\tilde{\varphi}_{n},e_{n}\right)\right) U^{*}=G\ltimes M.\qed
291: \end{displaymath}
292: \end{pf}
293:
294: \paragraph{ITPFI factors and Boca-Zaharescu factors}
295:
296: In \cite{ArW} Araki and Woods define ITPFI factors as infinite tensor product of type $\rm{I}$ factors
297: \begin{displaymath}
298: \bigotimes(M_{n},H_{n},\xi_{n}),
299: \end{displaymath}
300: where $M_{n}$ is a type $\rm{I}$ factor acting on $H_{n}$ and $\xi_{n}$ is a norm $1$ vector in $H_{n}$. All these factors are hyperfinite.
301:
302: If $M$ is a type $\rm{I}$ factor acting on $H$, we can write $H=H_{1}\otimes H_{2}$ such that $M=\mathcal{B}(H_{1})\otimes 1$. Now, let $\Omega\in H$ be a norm $1$ vector and consider the normal state on $M$
303: \begin{displaymath}
304: \omega(x)=\langle (x\otimes 1)\Omega,\Omega \rangle.
305: \end{displaymath}
306: Hence there exists a density matrix $\rho_{\Omega}\in\mathcal{B}(H_{1})$ such that $\omega(x)=Tr(\rho_{\Omega}x)$. It is easy to see that the ordered list (with multiplicity) of the non zero eigenvalues of the operator $\rho_{\Omega}$ does not depend on the decomposition of $H$ in $H_{1}\otimes H_{2}$. This list is denoted by Sp$(\Omega\,|\,M)$. The type of the ITPFI factor $\bigotimes\left(M_{n},H_{n},\xi_{n}\right)$ only depends on the list Sp$(\xi_{n}\,|\,M_{n})$. In the fourth section we will use the fact that if each $M_{n}$ is a type $\rm{I}_{n_{\nu}}$ factor, with $2\leq n_{\nu}\leq\infty$, and Sp$(\xi_{n}\,|\,M_{n})=\{\lambda_{n_{i}},\,i=1,2,\ldots,n_{\nu}\}$ then, if $\lambda_{n_{1}}\geq\delta$ for some $\delta>0$ and for all $n$, $M$ is a type $\rm{III}$ factor if and only if
307: \begin{displaymath}
308: \sum_{n,i}\lambda_{n_{i}}\inf\left\{\left|\frac{\lambda_{n_{1}}}{\lambda_{n_{i}}}-1\right|^{2},C\right\}=\infty,
309: \end{displaymath}
310: for some positive $C$.
311:
312: Let $\mathcal{S}$ be an infinite subset of the set $\mathcal{P}$ of all prime numbers and $\beta\in ]0,1]$. In \cite{Boca} Boca and Zaharescu studied the following ITPFI factor
313: \begin{displaymath}
314: M_{\beta,\mathcal{S}}:=\bigotimes_{p\in\mathcal{S}}\left(\mathcal{B}(l^{2}(\Nset)),\omega_{p,\beta}\right),
315: \end{displaymath}
316: where $\omega_{p}(x):=\sum_{n}p^{-n\beta}(1-p^{-\beta})\langle xe_{n},e_{n}\rangle$ and $(e_{n})$ is the canonical basis of $l^{2}(\Nset)$. We denote by $N_{\mathcal{S}}$ the factor $M_{1,\mathcal{S}}$. In \cite{Boca} Boca and Zaharescu show that
317: \begin{enumerate}
318: \item For any $\lambda\in [0,1]$ and $\beta\in ]0,1]$, there is a subset $\mathcal{S}$ of $\mathcal{P}$ such that $M_{\mathcal{\beta,S}}$ is a type $\rm{III}_{\lambda}$ factor.
319: \item For any $\beta\in ]0,1]$, any countable subgroup $K$ of $\Rset$ and any countable subset $\Sigma$ of $\Rset -K$, there exists a subset $\mathcal{S}$ of $\mathcal{P}$ such that $T(M_{\beta,\mathcal{S}})$ contains $K$ and does not intersect $\Sigma$,
320: \end{enumerate}
321: where $T(M)$ denotes the Connes' $T$ invariant of the von Neumann algebra $M$ (see \cite{Co1}).
322:
323: \begin{rem}
324: It was shown in \cite{Boca} that $M_{\beta,\mathcal{S}}$ is an $\text{ITPFI}_{2}$ (infinite tensor product of type $\rm{I}_{2}$ factors) for all $\beta\in ]\frac{1}{2},1]$. In fact it is possible to show that for all $\beta\in ]0,1]$, $M_{\beta,\mathcal{S}}$ is an $\text{ITPFI}_{m}$ with $\beta>\frac{1}{m}$. Indeed, for such $m$ and $\beta$ put
325: \begin{displaymath}
326: q_{p}(e_{n})=\left\{\begin{array}{cl}
327: e_{n} & \text{if $0\leq n \leq m-1$}\\
328: 0 & \text{otherwise.}\end{array}\right.
329: \end{displaymath}
330: Then because of
331: \begin{displaymath}
332: \sum_{p\in\mathcal{S}}\sum_{n\geq m}(1-p^{-\beta})p^{-n\beta}=\sum_{p\in\mathcal{S}}p^{-m\beta}<\infty,
333: \end{displaymath}
334: for all $\beta>\frac{1}{m}$, we can apply Lemma \ref{valprop} to obtain a projection $p\neq 0$ such that $p\left(M_{\beta,\mathcal{S}}\right)p$ is an $\text{ITPFI}_{m}$. Moreover, it is easy to see that $p$ is purely infinite, thus $M_{\beta,\mathcal{S}}$ is $\text{ITPFI}_{m}$.
335: \end{rem}
336: \section{The case of a finite factor}
337:
338: In this section we show that if $(M,\Delta)$ is a \lc{} quantum group such that $M$ is a finite factor, then $(M,\Delta)$ is compact so $\hat{M}$, being an infinite direct sum of full matrix algebras, is not a factor.
339:
340: The idea of the proof of the next lemma was taken from \cite{vanDaele}.
341: \begin{lem}
342: Let $(M,\Delta)$ be a \lc{} quantum group. Suppose that $M$ is a finite factor. Let $\tau$ be the unique tracial state on $M$. Then, for all $\rho\in M_{*}$ with $0\leq \rho\leq\tau$, one has :
343: \begin{displaymath}
344: \rho * \tau=\tau * \rho=\rho(1)\tau.
345: \end{displaymath}
346: \end{lem}
347:
348: \begin{pf}
349: Let $a$ be in $M$ and define $b=(\iota\otimes\tau)\Delta(a)$. Then, by unicity of $\tau$, one has $\tau * \tau = \tau$, and using the coassociativity of $\Delta$ we obtain
350: \begin{align*}
351: (\iota\otimes \tau)\Delta(b)
352: & =(\iota\otimes\tau)\Delta((\iota\otimes\tau)\Delta(a))
353: =(\iota\otimes\tau\otimes\tau)((\Delta\otimes\iota)\Delta(a)) \\
354: & =(\iota\otimes\tau\otimes\tau)((\iota\otimes\Delta)\Delta(a))\\
355: & =(\iota\otimes(\tau * \tau))\Delta(a)=(\iota\otimes\tau)\Delta(a)=b.
356: \end{align*}
357:
358: This implies the following relations.
359: \begin{align}
360: (\iota\otimes\tau)\left((b^{*}\otimes 1)\Delta(b)\right)
361: =b^{*}(\iota\otimes\tau)\Delta(b)=b^{*}b\label{eqtypII11}, \\
362: (\iota\otimes\tau)\left(\Delta(b^{*})(b\otimes 1)\right)
363: =\left((\iota\otimes 1\Delta(b)\right)^{*} b=b^{*}b.\label{eqtypeII12}
364: \end{align}
365:
366: Now define
367: \begin{align*}
368: k
369: & =\left(\Delta(b)-b\otimes 1\right)^{*}\left(\Delta(b)-b\otimes 1\right) \\
370: & =\Delta(b^{*}b)-(b^{*}\otimes 1)\Delta(b)-\Delta(b^{*})(b\otimes 1)+b^{*}b\otimes 1.
371: \end{align*}
372:
373: Then $k\geq 0$ and, from the equations (\ref{eqtypII11}) and (\ref{eqtypeII12}), we obtain
374: \begin{align*}
375: (\tau\otimes\tau)(k)
376: & =\tau((\iota\otimes\tau)(k))=(\tau\otimes\tau)(\Delta(b^{*}b)-\tau(b^{*}b) \\
377: & =\tau * \tau(b^{*}b)-\tau(b^{*}b)=0.
378: \end{align*}
379: Then, if $\rho\in M_{*}$ with $0\leq \rho\leq\tau$, one has $(\tau\otimes\rho)(k)\leq (\tau\otimes\tau)(k)=0$. This implies, with the Cauchy-Schwartz inequality, that for all $c\in M$ we have
380: \begin{displaymath}
381: (\tau\otimes\rho)\left((c\otimes 1)(\Delta(b)-b\otimes 1)\right)=0\qquad\text{thus},
382: \end{displaymath}
383: \begin{displaymath}
384: (\tau\otimes\rho)\left((c\otimes 1)(\Delta(b)\right)=\rho(1)\tau(cb).
385: \end{displaymath}
386: Using the definition of $b$, we see that the last equation is equivalent to
387: \begin{align*}
388: & (\tau\otimes\rho)\left((c\otimes 1)\Delta\left((\iota\otimes\tau)\Delta(a)\right)\right)=\rho(1)\tau\left(c(\iota\otimes\tau)\Delta(a)\right)\\
389: \Leftrightarrow
390: & (\tau\otimes\rho\otimes\tau)\left((c\otimes 1\otimes 1)(\Delta\otimes\iota)\Delta(a)\right)=\rho(1)(\tau\otimes\tau)\left((c\otimes 1)\Delta(a)\right)\\
391: \Leftrightarrow
392: & (\tau\otimes(\rho * \tau))\left((c\otimes 1)\Delta(a)\right)=\rho(1)(\tau\otimes\tau)\left((c\otimes 1)\Delta(a)\right),
393: \end{align*}
394: and this is true for all $a$ and $b$ in $M$. Now, because $\Delta(M)(M\otimes 1)$ is $\sigma$-weakly dense in $M\otimes M$ and $\tau$ is a trace we have, for all $x\in M\otimes M$,
395: \begin{displaymath}
396: (\rho(1)\tau\otimes\tau)(x)=(\tau\otimes(\rho * \tau))(x).
397: \end{displaymath}
398:
399: Putting $x=1\otimes y$ in the last equation, we obtain $\rho(1)\tau=\rho * \tau$. The proof of $\rho(1)\tau=\tau * \rho$ is the same.\qed
400: \end{pf}
401:
402: We are now able to prove that a \lc{} quantum group $(M,\Delta)$ with $M$ a finite factor is compact.
403:
404: \begin{thm}
405: Let $(M,\Delta)$ be a \lc{} quantum group with $M$ a finite factor. Then $(M,\Delta)$ is compact and $\tau$ is the Haar state on $M$, where $\tau$ is the unique tracial state on $M$.
406: \end{thm}
407:
408: \begin{pf}
409: Let $(H,\Lambda,\iota)$ be a G.N.S. construction for $\tau$ and $J$ the canonical involutive isometry associated to $\tau$. Let $a$ be in $M$ and consider the positive normal linear form $\omega_{\Lambda(a)}$. We have
410: \begin{align*}
411: \omega_{\Lambda(a)}(x^{*}x)
412: & =||\Lambda(xa)||^{2}=||Ja^{*}J\Lambda(x)||^{2} \\
413: & \leq ||a||^{2}\tau(x^{*}x).
414: \end{align*}
415: This implies that $\frac{\omega_{\Lambda(a)}}{||a||^{2}}\leq\tau$ and, using the previous lemma, we conclude that $\omega_{\Lambda(a)} * \tau=\tau * \omega_{\Lambda(a)}=\omega_{\Lambda(a)}(1)\tau$ for all $a\in M$. Now, using that $M\subset\mathcal{B}(H)$ is standard, we know that if $\omega\in M_{*}$ and $\omega\geq 0$ there exists $\xi\in H$ such that $\omega=\omega_{\xi}$. Take a net $(a_{i})$ in $M$ such that $\Lambda(a_{i})$ converges in $H$ to $\xi$ then, for all $x\in M$, $\omega_{\Lambda(a_{i})}(x)$ converges to $\omega(x)$. In particular, for $x$ in $M$, we have
416: \begin{align*}
417: (\omega_{\Lambda(a_{i})} * \tau)(x)\rightarrow (\omega * \tau)(x) \\
418: (\tau * \omega_{\Lambda(a_{i})})(x)\rightarrow(\tau * \omega)(x).
419: \end{align*}
420: Because of
421: \begin{displaymath}
422: (\omega_{\Lambda(a_{i})} * \tau)(x)=(\tau * \omega_{\Lambda(a_{i})})(x)=||\Lambda(a_{i})||^{2}\tau(x)\rightarrow ||\xi||^{2}\tau(x)=\omega(1)\tau(x),
423: \end{displaymath}
424: we see that $\omega * \tau=\tau * \omega=\omega(1)\tau$ and, by linearity, the last equality holds for all $\omega\in M_{*}$. This concludes the proof.\qed
425: \end{pf}
426:
427: \section{Examples}
428:
429: Let $\mathcal{P}$ be the set of all prime numbers. In the sequel, if
430: $p$ is a prime number, we denote by $\Qset_{p}$ the field of rational
431: $p$-adic numbers and $\Zset_{p}$ the ring of $p$-adic integers. Let
432: $\mathcal{S}$ be an infinite subset of $\mathcal{P}$ and
433: $\mathcal{A}_{\mathcal{S}}$ the restricted direct product of $\Qset_{p}$ relatively to the compact open subgroups $\Zset_{p}$ for
434: $p\in\mathcal{S}$ (see \cite{Bl}) :
435: \begin{displaymath}
436: \mathcal{A}_{\mathcal{S}}=\prod_{p\in\mathcal{S}}\,^{'}\left(\Qset_{p},\Zset_{p}\right).
437: \end{displaymath}
438: Then $\mathcal{A}_{\mathcal{S}}$ is a second countable \lc{} ring.
439: The group of invertible elements of $\mathcal{A}_{\mathcal{S}}$ is
440: \begin{displaymath}
441: \mathcal{A}^{*}_{\mathcal{S}}=\prod_{p\in\mathcal{S}}\,^{'}\left(\Qset_{p}^{*},\Zset_{p}^{*}\right).
442: \end{displaymath}
443: Now, denote by $G_{\mathcal{S}}$ the $ax+b$-group of
444: $\mathcal{A}_{\mathcal{S}}$ :
445: \begin{displaymath}
446: G_{\mathcal{S}}=\mathcal{A}^{*}_{\mathcal{S}}\ltimes\mathcal{A}_{\mathcal{S}},
447: \end{displaymath}
448: and define the following subgroups.
449:
450: \begin{displaymath}
451: G_{\mathcal{S}}^{1}=\left\{(a,0)\in G_{\mathcal{S}}\right\},\,\,\,
452: G_{\mathcal{S}}^{2}=\left\{\left((a_{p}),(b_{p})\right)\in
453: G,\,\,a_{p}+b_{p}p=1\,\,\forall p\in\mathcal{S}\right\}.
454: \end{displaymath}
455:
456: We can rewrite $G_{\mathcal{S}}^{2}$ as
457: \begin{displaymath}
458: G_{\mathcal{S}}^{2}=\left\{
459: \begin{array}{c}
460: \left((a_{p}),\left(\frac{1-a_{p}}{p}\right)\right),\,\,a_{p}\neq 0\,\,\,\forall p\in\mathcal{S}\\
461: \text{ and }a_{p}\in1+p\Zset_{p}\text{ for $p$ large enough}
462: \end{array}
463: \right\}.
464: \end{displaymath}
465: $G_{\mathcal{S}}^{1}$ is the subgroup of $G_{\mathcal{S}}$ which
466: fixes $0$. $G_{\mathcal{S}}^{2}$ is, formally, the subgroup of
467: $G_{\mathcal{S}}$ which fixes
468: $\left(\frac{1}{p}\right)_{p\in\mathcal{S}}$. We denote by
469: $\mu_{p}^{+}$ the additive Haar measure on $\Qset_{p}$ such that
470: $\mu_{p}^{+}(\Zset_{p})=1$ and by $\mu_{p}^{\times}$ the multiplicative
471: Haar measure on $\Qset_{p}^{*}$ such that $\mu_{p}^{\times}(\Zset_{p}^{*})=1$. Let
472: $\mu^{+}$ be the product measure of $\mu_{p}^{+}$, this is an
473: additive Haar measure on $\mathcal{A}_{\mathcal{S}}$, let
474: $\mu^{\times}$ be the product measure of $\mu_{p}^{\times}$, this is
475: a Haar measure on $\mathcal{A}^{*}_{\mathcal{S}}$. On
476: $G_{\mathcal{S}}$, the right Haar measure which is equal to $1$ on
477: $\prod_{p\in\mathcal{S}}\Zset_{p}^{*}\times\Zset_{p}$ is
478: $d\mu^{\times}(x)d\mu^{+}(y)$ and the left Haar measure which is
479: equal to $1$ on $\prod_{p\in\mathcal{S}}\Zset_{p}^{*}\times\Zset_{p}$ is
480: $\delta(x)d\mu^{\times}(x)d\mu^{+}(y)$, where
481: \begin{displaymath}
482: \delta(x)=\prod_{p\in\mathcal{S}} \frac{1}{|x_{p}|}_{p},\,\,\,x=(x_{p})\in\mathcal{A}_{\mathcal{S}}^{*}.
483: \end{displaymath}
484:
485: We now prove the following easy lemma.
486:
487: \begin{lem}
488: The groups $G_{\mathcal{S}}^{1}$, $G_{\mathcal{S}}^{2}$ are matched.
489: Moreover, the bicrossed product of $G_{\mathcal{S}}^{1}$ by
490: $G_{\mathcal{S}}^{2}$ is not regular, it is semi-regular in the
491: sense of \cite{BaaSkVa}.
492: \end{lem}
493:
494: \begin{pf}
495: It is clear that $G_{\mathcal{S}}^{1}$ and $G_{\mathcal{S}}^{2}$ are
496: closed subgroups of $G_{\mathcal{S}}$ and $G_{\mathcal{S}}^{1}\cap
497: G_{\mathcal{S}}^{2}=\{1\}$. So we must prove that
498: $G_{\mathcal{S}}-G_{\mathcal{S}}^{1}G_{\mathcal{S}}^{2}$ is closed
499: and its Haar measure is zero. From
500: \begin{displaymath}
501: G_{\mathcal{S}}^{2}G_{\mathcal{S}}^{1}=
502: \left\{
503: \begin{array}{c}
504: \left((a_{p}b_{p}),\left(\frac{1-a_{p}}{p}\right)\right),
505: \,\,b=(b_{p})_{p\in\mathcal{S}}\in\mathcal{A}^{*}_{\mathcal{S}},\,\,a_{p}\neq 0\,\,\,\forall p\in\mathcal{S}\\
506: \text{ and }a_{p}\in1+p\Zset_{p}\text{ for $p$ large enough}
507: \end{array}
508: \right\},
509: \end{displaymath}
510: we conclude that
511: \begin{displaymath}
512: G_{\mathcal{S}}^{2}G_{\mathcal{S}}^{1}=
513: \left\{(a,b)\in G_{\mathcal{S}},\,\,b_{p}\neq\frac{1}{p}\,\forall
514: p\in\mathcal{S}\text{ with }
515: b=(b_{p})_{p\in\mathcal{S}}\right\}.
516: \end{displaymath}
517: It follows that
518: $G_{\mathcal{S}}^{2}G_{\mathcal{S}}^{1}$ is open and
519: \begin{displaymath}
520: G_{\mathcal{S}}-G_{\mathcal{S}}^{2}G_{\mathcal{S}}^{1}=\left\{(a,b)\in G_{\mathcal{S}},\,\,\exists p\in \mathcal{S},\,\,
521: b_{p}=\frac{1}{p}\text{ with }b=(b_{p})_{p\in\mathcal{S}}\right\}
522: \end{displaymath}
523: has Haar measure equal to zero.\qed
524: \end{pf}
525:
526: Denote by $(M_{\mathcal{S}},\Delta_{\mathcal{S}})$ the bicrossed
527: product of $G_{\mathcal{S}}^{1}$ and $G_{\mathcal{S}}^{2}$. Under the canonical identification of $G_{\mathcal{S}}^{1}$ with
528: $\mathcal{A}^{*}_{\mathcal{S}}$ and $G_{\mathcal{S}}^{2}$ with
529: $\mathcal{K}_{\mathcal{S}}$, where $\mathcal{K}_{\mathcal{S}}$ is
530: the following restricted direct product
531: \begin{displaymath}
532: \mathcal{K}_{\mathcal{S}}=\prod_{p\in\mathcal{S}}\, ^{'}\left(\Qset_{p}^{*},1+p\Zset_{p}\right),
533: \end{displaymath}
534: the group actions $\alpha$ of $G_{\mathcal{S}}^{1}$ on the
535: measurable space $G_{\mathcal{S}}^{2}$ and $\beta$ of
536: $G_{\mathcal{S}}^{2}$ on the measurable space $G_{\mathcal{S}}^{1}$
537: can be easily calculated : take
538: $s=(s_{p})\in\mathcal{K}_{\mathcal{S}}$ and
539: $g=(g_{p})\in\mathcal{A}^{*}_{\mathcal{S}}$ such that for all
540: $p\in\mathcal{S}$, $g_{p}(s_{p}-1)+1\neq 0$ and, for $p$ large
541: enough, $g_{p}(s_{p}-1)+1\in 1+p\Zset_{p}$. Then
542: \begin{equation}\label{action}
543: \alpha_{g}(s)=\left(g_{p}(s_{p}-1)+1\right),\quad\beta_{s}(g)=\left(\frac{g_{p}s_{p}}{g_{p}(s_{p}-1)+1}\right).
544: \end{equation}
545: We define on $G_{\mathcal{S}}^{1}$ the Haar measure $\mu_{1}$
546: obtained, through the identification with
547: $\mathcal{A}^{*}_{\mathcal{S}}$, from the Haar measure
548: $\mu^{\times}$ on $\mathcal{A}^{*}_{\mathcal{S}}$. Also, we define
549: on $G_{\mathcal{S}}^{2}$ the Haar measure $\mu_{2}$ corresponding
550: to the product of the measures $\mu_{p}$ on $\Qset_{p}^{*}$, where $\mu_{p}$
551: is the Haar measure on $\Qset_{p}^{*}$ such that $\mu_{p}(1+p\Zset_{p})=1$. Taking
552: into account equation $(\ref{action})$, we see that $\alpha$ is a
553: restricted direct product action for $p\in\mathcal{S}$ of the
554: $\alpha^{p}\,:\,\Qset_{p}^{*}\rightarrow\text{Aut}(\Qset_{p}^{*})$,
555: $\alpha^{p}_{g_{p}}(s_{p})=g_{p}(s_{p}-1)+1$. Also $\beta$ is a
556: restricted direct product action of
557: $\beta^{p}\,:\,\Qset_{p}^{*}\rightarrow\text{Aut}(\Qset_{p}^{*})$,
558: $\beta^{p}_{s_{p}}(g_{p})=\frac{g_{p}s_{p}}{g_{p}(s_{p}-1)+1}$. We introduce the notation $\nu_{p}$ for the Haar measure on $\Qset_{p}$ such that $\nu_{p}(\Zset_{p}^{*})=1$. We have
559: \begin{displaymath}
560: \mu_{p}=(p-1)\mu_{p}^{\times}\, ,\,\,\nu_{p}=(1-p^{-1})^{-1}\mu_{p}^{+}\,\,\text{and}\,\, \d\mu_{p}^{+}(x)=(1-p^{-1})|x|_{p}\d\mu_{p}^{\times}(x).
561: \end{displaymath}
562:
563: The main result of this section is the following theorem which implies the description of the types of the factors $M_{\mathcal{S}}$ and $\hat{M}_{\mathcal{S}}$.
564:
565: \begin{thm}
566: For any infinite subset $\mathcal{S}$ of $\mathcal{P}$ we have
567: the following isomorphisms
568: \begin{displaymath}
569: M_{\mathcal{S}}\simeq
570: N_{\mathcal{S}}\qquad\text{and}\qquad\hat{M}_{\mathcal{S}}\simeq
571: M_{\mathcal{S}}\otimes\mathcal{R},
572: \end{displaymath}
573: where $N_{\mathcal{S}}$ is the
574: Boca-Zaharescu factor and $\mathcal{R}$ is the hyperfinite
575: $\rm{II}_{1}$ factor.
576: \end{thm}
577:
578: \begin{pf}
579: Let $\pi_{p}$ be the canonical inclusion of $L^{\infty}(\Qset_{p})$ in $\Qset_{p}^{*}\ltimes L^{\infty}(\Qset_{p})$. We first prove the following lemma.
580:
581: \begin{lem}\label{Vaes}
582: Let $\mu$ be a Haar measure on $\Qset_{p}^{*}$ and $\nu$ a Haar measure on
583: $\Qset_{p}$. Let $K\subset\Zset_{p}^{*}$ be a subgroup of finite index with
584: $\mu(K)=1$ and $L$ a compact open subset of $\Zset_{p}$ such
585: that $KL=L$ and $\nu(L)=1$. Define
586: \begin{displaymath}
587: e(K,L)=(\lambda(\11_{K})\otimes
588: 1)\pi_{p}(\11_{L})\qquad\text{and}\qquad\xi(K,L)=\11_{K\times L}.
589: \end{displaymath}
590: Then $e(K,L)$ is
591: a projection in $\Qset_{p}^{*}\ltimes L^{\infty}(\Qset_{p})$ and
592: \begin{displaymath}
593: \left(e(K,L)\left(\Qset_{p}^{*}\ltimes
594: L^{\infty}(\Qset_{p})\right)e(K,L),\omega_{\xi(K,L)}\right) \simeq
595: \left(\mathcal{B}(l^{2}(\Nset)),\omega\right)
596: \end{displaymath}
597: where $\omega$ is the
598: faithful normal state on $\mathcal{B}(l^{2}(\Nset))$ with eigenvalue
599: list given by
600: \begin{displaymath}
601: \frac{\mu_{p}^{+}(K)}{\mu_{p}^{+}(L)}p^{-n}\quad\text{with multiplicity}\quad\left|\quotient{L\cap
602: p^{n}\Zset_{p}^{*}}{K}\right|,\quad n\in\Nset.
603: \end{displaymath}
604: \end{lem}
605:
606: \begin{pf}
607: The fact that $e(K,L)$ is a projection has been proved in
608: Proposition \ref{crossed}. We define the following unitary
609: \begin{displaymath}
610: U\,\,:\,\,L^{2}(\Qset_{p}^{*}\times\Qset_{p},\mu\times\nu)\rightarrow
611: L^{2}(\Qset_{p}^{*}\times\Qset_{p}^{*},\mu\times\mu)
612: \end{displaymath}
613: \begin{displaymath}
614: (U\xi)(x,y)=\left(\frac{\mu_{p}^{+}(K)}{\mu_{p}^{+}(L)}|y|_{p})\right)^{\frac{1}{2}}\xi(xy^{-1},y).
615: \end{displaymath}
616: Then
617: \begin{displaymath}
618: U\lambda_{g}\otimes 1 U^{*}=\lambda_{g}\otimes
619: 1\quad\text{and}\quad U\pi_{p}(F) U^{*}=F\otimes 1,
620: \end{displaymath}
621: this implies
622: that
623: \begin{displaymath}
624: U\Qset_{p}^{*}\ltimes L^{\infty}(\Qset_{p})
625: U^{*}=\mathcal{B}(L^{2}(\Qset_{p}^{*},\mu))\otimes 1.
626: \end{displaymath}
627: Next, we have
628: \begin{displaymath}
629: (e(K,L)\xi)(x,y)=\11_{L}(xy)\int_{\Qset_{p}^{*}}\11_{K}(t)\xi(t^{-1}x,y)\d\mu(t)
630: \end{displaymath}
631: thus, after a simple computation, we obtain
632: \begin{displaymath}
633: Ue(K,L)U^{*}=f(K,L)\otimes 1\quad\text{where,}
634: \end{displaymath}
635: \begin{displaymath}
636: (f(K,L)\xi)(x)=\11_{L}(x)\int_{\Qset_{p}^{*}}\11_{K}(t)\xi(t^{-1}x)\d\mu(t).
637: \end{displaymath}
638: Observe that the image of $f(K,L)$ is the set of functions $\xi\in
639: L^{2}(\Qset_{p}^{*},\mu)$ such that the support of $\xi$ is in $L-\{0\}$ and
640: $\xi$ is invariant under translations of $K$. Writing
641: \begin{displaymath}
642: L-\{0\}=\cup_{n\in\Nset}L\cap p^{n}\Zset_{p}^{*},
643: \end{displaymath}
644: we see that every function $\xi$ in the image of $f(K,L)$ is of the
645: form
646: \begin{displaymath}
647: \xi=\sum_{n\in\Nset}\sum_{[y]\in\quotient{L\cap
648: p^{n}\Zset_{p}^{*}}{K}}\xi(y)\11_{[y]}.
649: \end{displaymath}
650: Thus we have
651: \begin{displaymath}
652: f(K,L)L^{2}(\Qset_{p}^{*},\mu)=\overline{\text{Span}<\11_{[y]},\,\,\,n\in\Nset,\,\,\,[y]\in\quotient{L\cap
653: p^{n}\Zset_{p}^{*}}{K}>},
654: \end{displaymath}
655: where $\overline{\text{Span}<X>}$ means the closed vector space generated by $X$. Because $\mu([y])=\mu(K)=1$, the set of vectors $\11_{[y]}$ for $[y]\in\quotient{L\cap
656: p^{n}\Zset_{p}^{*}}{K}$ and $n\geq 0$ is an orthonormal basis of $f(K,L)L^{2}(\Qset_{p}^{*},\mu)$. Thus, there is a unitary $W$ between $f(K,L)L^{2}(\Qset_{p}^{*},\mu)$ and $l^{2}(\Nset)$ such that
657: \begin{displaymath}
658: (W\otimes 1)Ue(K,L)\left(\Qset_{p}^{*}\ltimes L^{\infty}(\Qset_{p})\right)e(K,L)
659: U^{*}(W^{*}\otimes 1)=\mathcal{B}(l^{2}(\Nset))\otimes 1,
660: \end{displaymath}
661: and, using the computation
662: \begin{displaymath}
663: U\xi(K,L)=\sum_{n}\sum_{[y]\in\quotient{L\cap
664: p^{n}\Zset_{p}^{*}}{K}}\lambda_{n,[y]}^{\frac{1}{2}}\11_{[y]}\otimes\11_{[y]},
665: \end{displaymath}
666: where $\lambda_{n,[y]}=\frac{\mu_{p}^{+}(K)}{\mu_{p}^{+}(L)}p^{-n}$, we conclude the proof.\qed
667: \end{pf}
668:
669: \begin{rem}
670: We obtain, for $(K,L,\mu,\nu)=(\Zset_{p}^{*},\Zset_{p},\mu_{p}^{\times},\mu_{p}^{+})$, the list $(1-p^{-1})p^{-n}$ with multiplicity one and, for $(K,L,\mu,\nu)=(1+p\Zset_{p},\Zset_{p}^{*} -1,\mu_{p},\nu_{p})$, the following list : $(p-1)^{-1}$ with multiplicity $p-2$ and $(p-1)^{-1}p^{-n}$ with multiplicity $p-1$ for $n\geq 1$.
671: \end{rem}
672:
673: The next ingredient of the proof is the following lemma.
674:
675: \begin{lem}\label{Vaes2}
676: For any infinite subset $\mathcal{S}\subset\mathcal{P}$ we have
677: \begin{enumerate}
678: \item
679: $M_{\mathcal{S}}\simeq\bigotimes_{p\in\mathcal{S}}\left(\Qset_{p}^{*}\ltimes L^{\infty}(\Qset_{p}),L^{2}(\Qset_{p}^{*}\times\Qset_{p},\mu_{p}^{\times}\times\mu_{p}^{+}),\xi(\Zset_{p}^{*},\Zset_{p})\right),$
680: \item
681: $\hat{M}_{\mathcal{S}}\simeq\bigotimes_{p\in\mathcal{S}}\left(\Qset_{p}^{*}\ltimes L^{\infty}(\Qset_{p}),L^{2}(\Qset_{p}^{*}\times\Qset_{p},\mu_{p}\times\nu_{p}),\xi(1+p\Zset_{p},\Zset_{p}^{*}-1)\right).$
682: \end{enumerate}
683: \end{lem}
684:
685: \begin{pf}
686: To obtain the first isomorphism, recall that
687: \begin{displaymath}
688: G_{\mathcal{S}}^{1}\ltimes L^{\infty}(G_{\mathcal{S}}^{2})\simeq G_{\mathcal{S}}^{1}\ltimes L^{\infty}(\quotient{G_{\mathcal{S}}}{G_{\mathcal{S}}^{1}}),
689: \end{displaymath}
690: and because $G_{\mathcal{S}}^{1}=\mathcal{A}^{*}_{\mathcal{S}}\times\{0\}$, it is easy to see that
691: \begin{displaymath}
692: G_{\mathcal{S}}^{1}\ltimes L^{\infty}(G_{\mathcal{S}}^{2})\simeq\mathcal{A}^{*}_{\mathcal{S}}\ltimes L^{\infty}(\mathcal{A}_{\mathcal{S}}).
693: \end{displaymath}
694: Next, using Proposition \ref{crossed}, we obtain immediately the first isomorphism. For the second isomorphism, we first use Proposition \ref{crossed} and the discussion preceding the lemma to obtain
695: \begin{displaymath}
696: \hat{M}_{\mathcal{S}}\simeq
697: \bigotimes_{p\in\mathcal{S}}\left(\Qset_{p}^{*}\,_{\beta^{p}}\ltimes L^{\infty}(\Qset_{p}^{*}),L^{2}(\Qset_{p}^{*}\times\Qset_{p}^{*},\mu_{p}
698: \times\mu_{p}^{\times}),\11_{(1+p\Zset_{p})\times\Zset_{p}^{*}}\right).
699: \end{displaymath}
700: Now define
701: \begin{displaymath}
702: V\,:\,L^{2}(\Qset_{p}^{*}\times\Qset_{p}^{*},\mu_{p}\times\mu_{p}^{\times})\rightarrow
703: L^{2}(\Qset_{p}^{*}\times\Qset_{p},\mu_{p}\times\nu_{p}),
704: \end{displaymath}
705: \begin{displaymath}
706: (V\xi)(g,s)=\xi(g^{-1},(s+1)^{-1})|s+1|_{p}^{\frac{1}{2}}.
707: \end{displaymath}
708: $V$ is unitary and
709: \begin{displaymath}
710: V\left(\Qset_{p}^{*}\, _{\beta^{p}}\ltimes L^{\infty}(\Qset_{p}^{*})\right) V^{*}=\Qset_{p}^{*}\ltimes L^{\infty}(\Qset_{p}),
711: \end{displaymath}
712: where the action for the crossed product on the right is the translation. Finally, the computation
713: \begin{displaymath}
714: V\11_{(1+p\Zset_{p})\times\Zset_{p}^{*}}=\11_{(1+p\Zset_{p})\times(\Zset_{p}^{*}-1)}
715: \end{displaymath}
716: concludes the proof.\qed
717: \end{pf}
718:
719: We can now prove the Theorem. Using Lemmas \ref{Vaes} and \ref{Vaes2}, the remark between these two lemmas and Proposition \ref{qMq} we obtain, using the notation $e_{1}=\bigotimes_{p\in\mathcal{S}}e(\Zset_{p}^{*},\Zset_{p})$ and $e_{2}=\bigotimes_{p\in\mathcal{S}}e(1+p\Zset_{p},\Zset_{p}^{*} -1)$,
720: \begin{displaymath}
721: e_{1}M_{\mathcal{S}}e_{1}\simeq N_{S}\quad\text{and},\quad
722: e_{2}\hat{M}_{\mathcal{S}}e_{2}\simeq
723: \bigotimes_{p\in\mathcal{S}}\left(\mathcal{B}(l^{2}(\Nset)),\psi_{p}\right),
724: \end{displaymath}
725: where the eigenvalue list of $\psi_{p}$ is given by
726: \begin{displaymath}
727: (p-1)^{-1}p^{-n}\quad\text{with multiplicity}\quad
728: \left\{ \begin{array}{ll}
729: p-2 & \text{if $n=0$} \\
730: p-1 & \text{if $n\geq 1$.}
731: \end{array} \right.
732: \end{displaymath}
733: Next, because $M_{\mathcal{S}}$, $\hat{M}_{\mathcal{S}}$, $e_{1}$ and $e_{2}$ are purely infinite and $e_{1}$ and $e_{2}$ have central support equal to $1$ we have that $e_{1}M_{\mathcal{S}}e_{1}\simeq M_{\mathcal{S}}$ and $e_{2}\hat{M}_{\mathcal{S}}e_{2}\simeq\hat{M}_{\mathcal{S}}$. Thus, to conclude the proof, it is sufficient to prove that
734: \begin{displaymath}
735: \bigotimes_{p\in\mathcal{S}}\left(\mathcal{B}(l^{2}(\Nset)),\psi_{p}\right)\simeq N_{\mathcal{S}}\otimes\mathcal{R}.
736: \end{displaymath}
737: Using Lemma \ref{valprop} and
738: \begin{displaymath}
739: \sum_{p\in\mathcal{S}}\sum_{n\geq 1}(p-1)^{-1}p^{-n}=\sum_{p\in\mathcal{S}}(p-1)^{-2}<\infty
740: \end{displaymath}
741: we can remove one copy of $p^{-1},p^{-2},\ldots$ without changing the isomorphism class of the ITPFI factor (the projection obtained in Lemma \ref{valprop} is clearly purely infinite), thus, we obtain
742: \begin{displaymath}
743: \bigotimes_{p\in\mathcal{S}}\left(\mathcal{B}(l^{2}(\Nset)),\psi_{p}\right)\simeq\bigotimes_{p\in\mathcal{S}}\left(\mathcal{B}(l^{2}(\Nset))\otimes M_{p-2}(\Cset),\omega_{p}\otimes\tau_{p}\right),
744: \end{displaymath}
745: where $\tau_{p}$ is the normalized trace of the matrix algebra $M_{p-2}(\Cset)$. The theorem follows.\qed
746: \end{pf}
747:
748: \begin{cor}
749: For any infinite subset $\mathcal{S}\subset\mathcal{P}$, we have
750: \begin{enumerate}
751: \item $\sum_{p\in\mathcal{S}}\frac{1}{p}<+\infty \Leftrightarrow \mu^{+}(\mathcal{A}_{\mathcal{S}}-\mathcal{A}^{*}_{\mathcal{S}})=0 \Leftrightarrow (M_{\mathcal{S}},\Delta_{\mathcal{S}})$ is of type $(\rm{I}_{\infty},\rm{II}_{\infty})$.\label{cor1}
752: \item $\sum_{p\in\mathcal{S}}\frac{1}{p}=+\infty \Leftrightarrow \mu^{+}(\mathcal{A}^{*}_{\mathcal{S}})=0 \Leftrightarrow (M_{\mathcal{S}},\Delta_{\mathcal{S}})$ is of type $(\rm{III},\rm{III})$.
753: \end{enumerate}
754: Moreover we have
755: \begin{itemize}
756: \item For any $\lambda\in [0,1]$ there exists a subset $\mathcal{S}\subset\mathcal{P}$ such that $(M_{\mathcal{S}},\Delta_{\mathcal{S}})$ is of type $(\rm{III}_{\lambda},\rm{III}_{\lambda})$.
757: \item For any countable subgroup $K$ of $\Rset$ and countable subset $\Sigma$ of $\Rset -K$ there exists a subset $\mathcal{S}$ of $\mathcal{P}$ such that $T(M_{\mathcal{S}})$ contains $K$ and does not intersect $\Sigma$.
758: \end{itemize}
759: \end{cor}
760:
761: \begin{pf}
762: We have $\mu^{+}\left(\prod_{p\in\mathcal{S}}\Zset_{p}^{*}\right)=0 \Leftrightarrow
763: \mu^{+}(\mathcal{A}^{*}_{\mathcal{S}})=0$. Then, because
764: \begin{displaymath}
765: \mu^{+}\left(\prod_{p\in\mathcal{S}}\Zset_{p}^{*}\right)
766: =\prod_{p\in\mathcal{S}}\mu_{p}^{+}\left(\Zset_{p}^{*}\right)
767: =\prod_{p\in\mathcal{S}}\left(1-\frac{1}{p}\right),
768: \end{displaymath}
769: we have
770: \begin{equation}\label{equiv}
771: \sum_{p\in\mathcal{S}}\frac{1}{p}=+\infty \Leftrightarrow
772: \mu^{+}(\mathcal{A}^{*}_{\mathcal{S}})=0.
773: \end{equation}
774: Now, the Borel-Cantelli lemma gives
775: $\sum_{p\in\mathcal{S}}\frac{1}{p}<+\infty \Rightarrow
776: \mu^{+}(\mathcal{A}_{\mathcal{S}}-\mathcal{A}^{*}_{\mathcal{S}})=0,$
777: then $(\ref{equiv})$ implies that the last implication is an
778: equivalence. Note that for any \lc{} ring $\mathcal{A}$, such that
779: $\mathcal{A}-\mathcal{A}^{*}$ has additive Haar measure zero, the
780: translation action of $\mathcal{A}^{*}$ on $\mathcal{A}$ is free and
781: ergodic, and the corresponding crossed product is a type
782: $\rm{I}_{\infty}$ factor, the proof of $(\ref{cor1})$ follows.
783:
784: Now, suppose that
785: $\sum_{p\in\mathcal{S}}\frac{1}{p}=\infty$ then
786: \begin{displaymath}
787: \sum_{p\in\mathcal{S},i\geq
788: 0}p^{-i}(1-p^{-1})\inf\{|p^{i}-1|^{2},1\}=
789: \sum_{p\in\mathcal{S},i\geq
790: 1}p^{-i}(1-p^{-1})=\sum_{p\in\mathcal{S}}\frac{1}{p}=+\infty.
791: \end{displaymath}
792: This implies, taking into account preliminaries about ITPFI factors, that $M_{\mathcal{S}}$ and $\hat{M}_{\mathcal{S}}$ are type
793: $\rm{III}$ factors. The last results follow from \cite{Boca}.\qed
794: \end{pf}
795:
796: There is a minor modification of the preceding example. Take
797: \begin{displaymath}
798: G_{\mathcal{S}}=\mathcal{K}_{\mathcal{S}}\ltimes\mathcal{A}_{\mathcal{S}}
799: \end{displaymath}
800: and define the following subgroups
801: \begin{displaymath}
802: G^{1}_{\mathcal{S}}=\mathcal{K}_{\mathcal{S}}\times\{0\}\quad\text{and}\quad G^{2}_{\mathcal{S}}=\left\{\left(a_{p},\frac{1-a_{p}}{p}\right),\,(a_{p})\in\mathcal{K}_{\mathcal{S}}\right\}.
803: \end{displaymath}
804: Then it is easy to see that $(G^{1}_{\mathcal{S}},G^{2}_{\mathcal{S}})$ is a matched pair. A direct computation gives, for $(a_{p}),\,(b_{p})\in\mathcal{K}_{\mathcal{S}}$,
805: \begin{displaymath}
806: \alpha_{(a_{p},0)}\left(b_{p},\frac{1-b_{p}}{p}\right)
807: =\left(a_{p}(b_{p}-1)+1,\frac{a_{p}(1-b_{p})}{p}\right)\quad
808: \text{and,}
809: \end{displaymath}
810: \begin{displaymath}
811: \beta_{\left(b_{p},\frac{1-b_{p}}{p}\right)}(a_{p},0)
812: =\left(\frac{a_{p}b_{p}}{a_{p}(b_{p}-1)+1},0\right).
813: \end{displaymath}
814: We can construct the bicrossed product \lc{} quantum group $(L_{\mathcal{S}},\Delta_{\mathcal{S}})$ having the following property.
815:
816: \begin{prop}
817: For any infinite subset $\mathcal{S}$ of $\mathcal{P}$, the \lc{} quantum group $(L_{\mathcal{S}},\Delta_{\mathcal{S}})$ is self-dual and
818: \begin{displaymath}
819: L_{\mathcal{S}}\simeq N_{\mathcal{S}}\otimes \mathcal{R}.
820: \end{displaymath}
821: \end{prop}
822:
823: \begin{pf}
824: Define the isomorphism $u\,:\,G^{1}_{\mathcal{S}}\rightarrow G^{1}_{\mathcal{S}}$ by
825: \begin{displaymath}
826: u(a_{p},0)=\left(a_{p}^{-1},\frac{1-a_{p}^{-1}}{p}\right),
827: \end{displaymath}
828: one verifies that
829: \begin{displaymath}
830: u\left(\beta_{\left(b_{p},\frac{1-b_{p}}{p}\right)}(a_{p},0)\right)
831: =\alpha_{u^{-1}\left(b_{p},\frac{1-b_{p}}{p}\right)}\left(u(a_{p},0)\right).
832: \end{displaymath}
833: Hence, interchanging $\alpha$ and $\beta$, we get an isomorphic matched pair and so an isomorphic \lc{} quantum group. To obtain the isomorphism, recall that $G_{\mathcal{S}}^{1}\ltimes L^{\infty}(G_{\mathcal{S}}^{2})\simeq G_{\mathcal{S}}^{1}\ltimes L^{\infty}(\quotient{G_{\mathcal{S}}}{G_{\mathcal{S}}^{1}})$, and because $G_{\mathcal{S}}^{1}=\mathcal{K}_{\mathcal{S}}\times\{0\}$, it is easy to see that $G_{\mathcal{S}}^{1}\ltimes L^{\infty}(G_{\mathcal{S}}^{2})\simeq\mathcal{K}_{\mathcal{S}}\ltimes L^{\infty}(\mathcal{A}_{\mathcal{S}})$. Next, using Lemma \ref{crossed} we obtain
834: \begin{displaymath}
835: L_{\mathcal{S}}\simeq\bigotimes_{p\in\mathcal{S}}\left(\Qset_{p}^{*}\ltimes L^{\infty}(\Qset_{p}),L^{2}(\Qset_{p}^{*}\times\Qset_{p},\mu_{p}\times\mu_{p}^{+}),\xi(1+p\Zset_{p},\Zset_{p})\right).
836: \end{displaymath}
837: This implies, using Lemma \ref{Vaes} with $(K,L,\mu,\nu)=(1+p\Zset_{p},\Zset_{p},\mu_{p},\mu_{p}^{+})$ and Proposition \ref{qMq}, the following isomorphism
838: \begin{displaymath}
839: L_{\mathcal{S}}\simeq\bigotimes_{p\in\mathcal{S}}\left(\mathcal{B}(l^{2}(\Nset))\otimes M_{p-1}(\Cset),\omega_{p}\otimes\tau_{p}\right),
840: \end{displaymath}
841: where $\tau_{p}$ is the normalized trace of the matrix algebra $M_{p-1}(\Cset)$.\qed
842: \end{pf}
843:
844: \begin{ack}
845: I would like to thank Prof. S. Vaes for suggesting the subject of the paper and for many comments and Prof. F. Boca for helpful discussions. I am also grateful to Prof. L. Vainerman for many helpful discussions and comments.
846: \end{ack}
847:
848: \bibliographystyle{plain}
849: \bibliography{PFima.bib}
850:
851: \end{document}
852: