1: %&biglatex

2: \documentclass[a4paper,11pt,leqno]{amsart}

3: \usepackage[latin1]{inputenc}

4: %\usepackage[T1]{fontenc}

5: %\usepackage{babel}

6: \usepackage{amsmath} 

7: \usepackage{mathtools} 

8: \usepackage{amssymb}

9: \usepackage{enumerate} 

10: \usepackage{xspace} 

11: \usepackage{amsfonts}

12: \usepackage{latexsym} 

13: \usepackage{amsthm} 

14: \usepackage[backrefs]{amsrefs}

15: \usepackage[all]{xypic} 

16: %\usepackage{pdfsync}

17: \usepackage{hyperref}

18: 

19: \numberwithin{equation}{section}

20: %\multlinegap2cm

21: 

22: \newtheorem{lemma}{Lemma}[section]

23: \newtheorem{proposition}[lemma]{Proposition}

24: \newtheorem{theorem}[lemma]{Theorem}

25: \newtheorem{corollary}[lemma]{Corollary}

26: \newtheorem{claim}{Claim}

27: \newtheorem*{clai}{Claim}

28: 

29: \theoremstyle{definition}

30: \newtheorem{definition}[lemma]{Definition}

31: \newtheorem{example}[lemma]{Example}

32: \newtheorem*{exam}{Example}

33: \newtheorem{remark}[lemma]{Remark}

34: 

35: \def\C{\mathbb{C}}

36: \def\N{\mathbb{N}}

37: \def\No{{\N_0}}

38: \def\Z{\mathbb{Z}}

39: \def\R{\mathbb{R}}

40: \def\O{\mathcal{O}}

41: \def\D{\widetilde{D}}

42: \def\t{\widetilde{\theta}}

43: \def\X{\hat{X}}

44: \def\H{\mathsf{H}}

45: \def\A{\ba{X}}

46: \def\Aa{\mathcal{A}}

47: \def\DX{\mathcal{D}_{\OSS}}

48: 

49: \DeclareMathOperator{\proj}{proj}

50: \DeclareMathOperator{\spa}{span}

51: \DeclareMathOperator{\Id}{Id}

52: 

53: \newcommand{\pxa}{\bigl(\XA,\theta_\XA,\fg{\ind}\bigr)}

54: \newcommand{\cspxa}{\cs\bigl(\XA,\theta_\XA,\fg\ind\bigr)}

55: \newcommand{\xanp}{\bigl(\XA,\theta_\XA,\fg{n}\bigr)}

56: \newcommand{\K}{\mathbf{K}}

57: \newcommand{\indd}{\N^2_0}

58: \newcommand{\lee}{\preceq}

59: \newcommand{\gee}{\succeq}

60: \newcommand{\BO}{\Bx(\O_A)}

61: \newcommand{\Bx}{\mathcal{B}}

62: \newcommand{\B}{\Bx(X,\theta,G)}

63: \newcommand{\JJJ}{\mathcal{J}}

64: \newcommand{\cy}[2]{C(#1,#2)}

65: \newcommand{\spc}{\overline{\spa}}

66: \newcommand{\coni}{(*)}

67: \newcommand{\conii}{(**)}

68: \newcommand{\bb}[1]{\mathbf{#1}}

69: \newcommand{\cs}{C^*}

70: \newcommand{\inv}{^{-1}}

71: \newcommand{\ba}[1]{\bar{#1}}

72: \newcommand{\h}[1]{\hat{#1}}

73: \newcommand{\fg}[1]{\mathbb{F}_{#1}}

74: \newcommand{\al}{\mathfrak a}

75: \newcommand{\Past}{{\mathcal P}}

76: \newcommand{\equ}[2]{\hspace{2pt}_{#1}\hspace{-3pt}\sim_{#2}}

77: \newcommand{\eq}{\equ{k}{l}}

78: \newcommand{\OSS}[1][]{{\mathsf{X}^{+}_{#1}}}

79: \newcommand{\TSS}{\mathsf{X}}

80: \newcommand{\MSS}{\mathsf{X}}

81: \newcommand{\laTSS}{\mathcal{L}(\TSS)}

82: \newcommand{\laOSS}{\mathcal{L}(\OSS)}

83: \newcommand{\ind}{\mathcal{I}}

84: \newcommand{\ec}[3]{{}_{#1}\hspace{-1pt}[#2]_{#3}}

85: \newcommand{\osh}{\sigma}

86: \newcommand{\tsh}{\tau}

87: \newcommand{\msh}{\sigma}

88: \newcommand{\alwords}{\al^*}

89: \newcommand{\XA}{{\OSS[A]}}

90: \newcommand{\tsha}{\osh}

91: \newcommand{\kgraph}{\Lambda}

92: \newcommand{\Dx}{\mathcal{D}_{\OSS}}

93: \newcommand{\BA}{\Bx(\XA,\theta_\XA,\fg\ind)}

94: \newcommand{\Bk}{\Bx_{\kgraph}}

95: \newcommand{\XX}{\mathcal{X}}

96: \newcommand{\cset}{\mathcal{F}}

97: \newcommand{\boomap}[1][(X,\theta,G)]{\phi_{#1}}

98: \newcommand{\bound}{\mathsf{B}}

99: \newcommand{\OSD}{\mathrm{X^+}}

100: \newcommand{\osd}{\sigma}

101: \newcommand{\TSD}{\mathrm{X}}

102: \newcommand{\tsd}{\tau}

103: \newcommand{\csp}[1]{\cs(#1)}

104: \newcommand{\csof}[1]{\cs\left(\left\{#1\right\}\right)}

105: \newcommand{\BXY}{\Bx(X\setminus Y,\theta_{|X\setminus Y},G)}

106: \newcommand{\crosmap}{\hspace{.3ex}\eta_\rtimes}

107: \newcommand{\tcrosmap}{\tilde{\eta}_\rtimes}

108: \newcommand{\fw}{\mathcal{W}}

109: \newcommand{\find}{\mathcal{F}}

110: \newcommand{\ideal}{\mathcal{I}}

111: \newcommand{\uset}[1][\ideal]{U(#1)}

112: \newcommand{\ioss}{\eta_\O}

113: \newcommand{\pnes}{\mathcal{E}}

114: \newcommand{\spro}{\pi}

115: \newcommand{\boof}{\eta}

116: \newcommand{\nTSS}{n_\TSS}

117: \newcommand{\emptyword}{\varepsilon}

118: \newcommand{\rte}{\mathcal{J}_\TSS}

119: \newcommand{\jj}{\mathbf{j}}

120: \newcommand{\xx}{\mathbf{x}}

121: 

122: \renewcommand{\mu}{u}

123: \renewcommand{\nu}{v}

124: 

125: \providecommand{\norm}[1]{[#1]} 

126: \providecommand{\abs}[1]{\lvert#1\rvert}

127: 

128: \begin{document}

129: \title[Dynamics, Boolean algebras and $\cs$-algebras]{Symbolic

130:   dynamics, partial dynamical systems, Boolean algebras and

131:   $\cs$-algebras generated by partial isometries}

132: \author{Toke Meier Carlsen}

133: \date{\today}

134: \address{Mathematisches Institut\\

135: Einsteinstra\ss e 62\\

136: 48149 M\" unster\\

137: Germany}

138: \thanks{This research has been supported by the EU-Network Quantum

139:   Spaces - Noncommutative Geometry (HPRN-CT-2002-00280).} 

140: \curraddr{Department of Mathematics\\

141:   University of Newcastle\\

142:   NSW 2308\\

143:   Australia}

144: \email{toke@math.uni-muenster.de} 

145: \keywords{$C^*$-algebras, partial dynamical systems, symbolic

146:   dynamical systems, Boolean algebras}

147: \subjclass[2000]{Primary 46L55; Secondary 46L05, 37B10, 06E99}

148: 

149: \begin{abstract}

150:   We associate to each discrete partial dynamical system a universal

151:   $\cs$-algebra generated by partial isometries satisfying relations

152:   given by a Boolean algebra connected to the discrete partial

153:   dynamical system in question. We show that for symbolic dynamical

154:   systems like one-sided and two-sided shift spaces and topological

155:   Markov chains with an arbitrary state space the $\cs$-algebras

156:   usually associated to them can be obtained in this way.

157: 

158:   As a consequence of this, we will be able to show that for two-sided

159:   shift spaces having a certain property, the crossed product of the

160:   two-sided shift space is a quotient of the $\cs$-algebra associated

161:   to the corresponding one-sided shift space.

162:  \end{abstract}

163: 

164: \maketitle

165: 

166: \section{Introduction}

167: 

168: The history of associating $\cs$-algebras to symbolic dynamical

169: systems is long and successful.  

170: 

171: A good example of this is the crossed product of infinite minimal

172: two-sided shift spaces which in \cite{MR1363826} was used to classify infinite

173: minimal shift spaces up till strong orbit equivalence and flip

174: conjugacy (it is actually done for a bigger class of dynamical

175: systems, namely Cantor systems, but we will in this paper only concern

176: ourselves with symbolic dynamical systems). 

177: 

178: Another very important example of a class of $\cs$-algebras associated

179: to symbolic dynamical systems is the Cuntz-Krieger algebras

180: \cite{MR561974}, which in a natural way can be viewed as

181: $\cs$-algebras associated to topological Markov chains with finite

182: state space. The Cuntz-Krieger algebras have proved to be very

183: important examples in the theory of $\cs$-algebras and have also let

184: to invariants of shift of finite type such as the dimension group

185: (cf. \cite{MR561973} and \cite{MR561974}). 

186: 

187: The Cuntz-Krieger algebras have been generalized in many different

188: ways. We will in this paper focus on two of those. The first one is

189: due to Exel and Lace, who in \cite{MR2000i:46064} have generalized the

190: Cuntz-Krieger algebras to topological Markov chains with arbitrary

191: state space. The other generalization is due to Matsumoto,

192: who in \cite{MR1454478} associated a $\cs$-algebra to every shift

193: space (called a subshift in that paper). Matsumoto's original

194: construction associated a $\cs$-algebra to every \emph{two-sided}

195: shift space, but it is more natural to view it as a way to associated

196: a $\cs$-algebra to every \emph{one-sided} shift space (cf. \cite{tmc}

197: and \cite{CS}).  

198: 

199: Topological Markov chains with finite state space are examples of

200: one-sided shift spaces, and it turns out that the $\cs$-algebras

201: associated to these kind of shift spaces are Cuntz-Krieger algebras,

202: so in this way the class of $\cs$-algebras associated with shift

203: spaces is a generalization of the class of Cuntz-Krieger algebras

204: (cf. \cite{CS}*{Section 8}).  

205: 

206: Thus we have three different classes of $\cs$-algebras associated to

207: symbolic dynamical system, namely crossed products of two-sided shift

208: spaces, Exel and Laca's generalization of Cuntz-Krieger algebras and

209: $\cs$-algebras associated to one-sided shift spaces. The main purpose

210: of this paper is to unify these three constructions to one.

211:  

212: My original motivation for written this paper was to prove Theorem

213: \ref{theorem:toke}, which for shift spaces having a certain property

214: relates the crossed product of the two-sided shift space and the

215: $\cs$-algebra associated to the corresponding one-sided shift

216: space. Doing this I found that the crossed product of two-sided shift

217: spaces and the $\cs$-algebra associated to one-sided shift spaces have

218: a common structure, which I also found in Exel and Laca's

219: generalization of Cuntz-Krieger algebras (which we for simplicity from

220: now on just will call Cuntz-Krieger algebras). This structure can be

221: described by partial representations of groups and Boolean

222: algebras. More formal, what we will do is to associate to every so

223: called discrete partial dynamical system a $\cs$-algebra and then to

224: every symbolic dynamical system (both one-sided and two-sided)

225: associate a discrete partial dynamical system in such a way that the

226: $\cs$-algebras we get in this way for one-sided shift spaces,

227: two-sided shift spaces and topological Markov chains are canonical

228: isomorphic to the $\cs$-algebra associated to the one-sided shift

229: space, the crossed product of the two-sided shift space and the

230: unitization of the Cuntz-Krieger of the transition matrix of the

231: topological Markov chain, respectively.

232: 

233: This construction is very natural, and I hope that beside the

234: benefits from having a unified construction of these different classes

235: of $\cs$-algebras associated to different kinds of symbolic dynamical

236: systems, the construction will also clarify in which way the

237: structure of the symbolic dynamical system is reflected in the

238: structure of the associated $\cs$-algebra.

239: 

240: The paper is organized as follows: In Section \ref{sec:notat}, we will

241: shortly introduce some notation which will be used throughout the

242: paper, in Section \ref{ekse} \emph{discrete partial dynamical

243:   systems} will be defined, and we will see how we from one- and

244: two-sided symbolic dynamical systems can construct discrete partial

245: dynamical systems. We will then in Section \ref{sec:cs-algebra} define

246: the $\cs$-algebra of a discrete partial dynamical system and show some

247: basic properties of it. In Section \ref{sec:construction} we will show

248: the $\cs$-algebra of a discrete partial dynamical system can be

249: constructed as a crossed product of a $\cs$-partial dynamical systems,

250: and we will in Section \ref{sec:repr} construct a representation of

251: the $\cs$-algebra of a discrete partial dynamical system as operators

252: on a Hilbert space. Section \ref{retu} is the main section of this

253: paper; here we will show that the crossed product of a two-sided shift

254: space, the $\cs$-algebra of a one-sided shift space, and Cuntz-Krieger

255: algebras can be obtained as $\cs$-algebras of discrete partial

256: dynamical systems. We will then in Section \ref{sec:ideal} describe

257: the ideal structure of the $\cs$-algebra of a discrete partial

258: dynamical system and then use this description to prove the above

259: mentioned Theorem \ref{theorem:toke}. The paper finish with three

260: appendices in which we will give a short introduction to partial

261: representations of groups, Boolean algebras and crossed products of

262: $\cs$-partial dynamical systems.

263: 

264: A previous version of this paper appeared in my Ph.D thesis

265: \cite{phd}. Unfortunately that version contained a lot of mistakes, which

266: hopefully have been fixed in this version. The most notable of these

267: mistakes was that I claimed that the $\cs$-algebra of a higher rank graph

268: could be constructed as the $\cs$-algebra of a discrete partial

269: dynamical system. The proof of this is however false, but it is

270: possible to construct the $\cs$-algebras of a higher rank graph

271: in a very similar way by using an action of an (discrete) inverse

272: semigroup instead of a partial action of a (discrete) group. This will

273: be proved in a forthcoming paper by Gwion Evans and the author.

274: 

275: \subsection*{Acknowledgment}

276: The process of writing this paper has been very long. I started on it

277: will I was a Ph.D-student at the University of Copenhagen, I continued

278: working on it while I was a post Doc at the Institut Mittag-Leffler and

279: at the Norwegian University of Science and Technology and I finally

280: finished it as a post Doc at Universit\" at M\" unster. I wish to

281: thank all members of the operator algebra groups at these places

282: for their kind hospitality and especially S\o ren Eilers, Christian Skau

283: and Joachim Cuntz. I also wish to thank Aidan Sims for pointing out

284: the above mentioned mistake. 

285: 

286: \section{Notation and preliminaries} \label{sec:notat}

287: Throughout this paper, $e$ will denote the \emph{neutral element} of a

288: given group. We will by $\Z$ denote \emph{the set of integers}, $\N_0$

289: will denote \emph{the set of non-negative integers} and $\N$ will

290: denote \emph{the set of positive integers}. 

291: 

292: If $X$ is a set, then we will by $\Id_X$ denote \emph{the identity

293: map on $X$}. For a map $\sigma:X\to X$, we will for every $k\in\N$ by

294: $\sigma^k$ denote \emph{the $k$-times composition of $\sigma$ with itself},

295: and we will set $\sigma^0=\Id_X$. If $\sigma$ is invertible, then we

296: will for every $k\in\N$ by $\sigma^{-k}$ denote the map

297: $(\sigma\inv)^k$. If $A$ is a subset of $X$, then $1_A$ denotes \emph{the

298: characteristic function}:

299: \begin{equation*}

300:   1_A(x)=

301:   \begin{cases}

302:     1&\text{if }x\in A,\\

303:     0&\text{if }x\notin A.

304:   \end{cases}

305: \end{equation*}

306: If $\theta$ is a map defined on $A$, then we will for another subset

307: $B$ of $X$ by $\theta(B)$ mean $\theta(A\cap B)$, and we will by

308: $\theta_{|B}$ denote the restriction of $\theta$ to $A\cap B$. 

309: 

310: If $C$ is a subset of a vector space, then we will use $\spa(C)$ to

311: denote \emph{the linear span of $C$}, and if the vector space also comes with a

312: topology, then $\spc(C)$ will denote \emph{the closure of the linear

313:   span of $C$}. For a subset $D$ of a $\cs$-algebra we denote

314: \emph{the $\cs$-subalgebra generated by $D$}, by $\cs(D)$.

315: 

316: When $\al$ is a set, then we will by $\fg\al$ denote \emph{the free group

317: generated by $\al$}. We will regard $\al$ as a subset of $\fg\al$ and

318: denote the subset $\bigl\{a\inv\mid a\in\al\bigr\}$ of $\fg\al$ by

319: $\al\inv$. We say that an element $g\in\fg\al$ \emph{is written

320: in reduced form $b_1b_2\dotsm b_k$} if $b_1,b_2,\dotsc,b_k\in\al\cup\al\inv$ and

321: $b_j=b\Rightarrow b_{j+1}\ne b\inv$ for every $j\in\{1,2,\dotsc,k-1\}$

322: and every $b\in\al\cup\al\inv$. We let $[\cdot]$ be the unique

323: homomorphism from $\fg\al$ to $\Z$ such that $[a]=1$ for every $a\in\al$. 

324: 

325: We will by $\al^*$ denote \emph{the set of finite words of symbols from

326: $\al$}. If $\mu\in \al^*$, then we will by $\abs{\mu}$ denote \emph{the

327: length of $\mu$} (i.e., the number of symbols in $\mu$), by $\mu_1$ the

328: first letter (the leftmost) letter of $\mu$, by $\mu_2$ the second

329: letter of $\mu$, and so on till $\mu_{|\mu|}$ which denotes the last

330: (the rightmost) letter of $\mu$. Thus $\mu=\mu_1\mu_2\dotsm

331: \mu_{|\mu|}$. We will identify $\al^*$ with the positive cone of $\fg\al$ 

332: which is defined to be the unital sub-semigroup of $\fg\al$ generated by

333: $\al$. Under this identification, the unit element of $\fg\al$ is

334: equal to the \emph{empty word}, which we will denote by $\emptyword$,

335: and $\abs{\mu}$ is equal to $\norm{\mu}$ for all $\mu\in\al^*$.

336: 

337: We will denote the set of one-sided, respectively two-sided, infinite

338: sequences in $\al$ by $\al^\No$, respectively $\al^\Z$. We will

339: often denote an element $x=(x_n)_{n\in\N}$ of $\al^\No$ by

340: \begin{equation*}

341:   x_0x_1\dotsm ,

342: \end{equation*}

343: and if $\mu\in\al^*$, then we will by $\mu x$ denote the sequence

344: \begin{equation*}

345:   \mu_1\mu_2\dotsm \mu_{|\mu|}x_0x_1\dotsm .

346: \end{equation*}

347: We will also often for a sequence $x$ belonging to either $\al^\No$

348: or $\al^\Z$ and for integers $k< l$ belonging to the appropriate

349: index set denote $x_kx_{k+1}\dotsm x_{l-1}$ by $x_{[k,l[}$ and regard

350: it as an element of $\al^*$. Likewise will $x_{[k,\infty[}$ denote

351: the element

352: \begin{equation*}

353:   x_kx_{k+1}\dotsm

354: \end{equation*}

355: of $\al^\N$.

356: 

357: We say that an element $(x_n)_{n\in\Z}\in\al^\Z$ is \emph{periodic} if

358: there exists an $m\in\N$ such that $x_{n+m}=x_n$ for all $n\in\Z$, and

359: we say that an element $(x_n)_{n\in\No}\in\al^{\No}$ is \emph{eventually

360:   periodic} if there exist $m,N\in\N$ such that $x_{n+m}=x_n$ for $n>N$.

361: 

362: \section{Discrete partial dynamical systems} \label{ekse}

363: We are now going to define \emph{partial actions} and \emph{discrete

364:   partial dynamical systems} and look at some ways to construct

365: discrete partial dynamical systems. 

366: 

367: Partial actions have been defined and studied by Ruy Exel in

368: \cite{MR1469405}, where he for any given group constructed an inverse

369: semigroup such that there is a one-to-one correspondence between the actions of

370: the inverse semigroup and the partial actions of the group.

371: 

372: \begin{definition}

373:   Given a group $G$ and a set $X$, a \emph{partial action} $\theta$ of

374:   $G$ on $X$ is a pair

375:   \begin{equation*}

376:     \bigl((D_g)_{g\in G},(\theta_g)_{g\in G}\bigr),

377:   \end{equation*}

378:   where for each $g\in G$, $D_g$ is a subset of $X$ and $\theta_g$ is

379:   a bijective map from $D_{g\inv}$ to $D_g$, satisfying for all $h$

380:   and $i$ in $G$:

381:   \begin{subequations}    

382:     \begin{align}

383:       &D_e=X \text{ and } \theta_e \text{ is the identity map on }X,\\

384:       &\theta_h(D_i)=D_h\cap D_{hi}, \label{eq:paractb}\\

385:       &\theta_h(\theta_i(x))=\theta_{hi}(x) \text{ for }x\in

386:       D_{i\inv}\cap D_{i\inv h\inv}.

387:     \end{align}

388:   \end{subequations}

389:   We will call the family $(D_g)_{g\in G}$ the \emph{domains} of

390:   $\theta$ and the family $(\theta_g)_{g\in G}$ the \emph{partial

391:     one-to-one maps} of $\theta$.

392: 

393:   The triple $(X,\theta,G)$ is called a \emph{discrete partial

394:     dynamical system}.

395: \end{definition}

396: 

397: The reason that we have chosen to call such a triple a \emph{discrete}

398: partial dynamical system is that we do not require any structure of

399: the system $(X,\theta,G)$ other than the above mentioned. It is in many

400: cases (see for example \cite{MR2003f:46108}) natural to ask for the set

401: $X$ to be a topological space, the domains $(D_g)_{g\in G}$ open

402: subsets, and the maps $(\theta_g)_{g\in G}$ homeomorphisms, but we

403: will in this paper only consider the discrete case, where we do not

404: require such a structure. 

405: 

406: A very simple example of a discrete partial dynamical systems is if we for a

407: given group $G$ and a given set $X$ for every $g\in G$ let $D_g=X$ and

408: let $\theta_g$ be the identity map on $X$. We get slightly more

409: interesting examples if we consider group actions:

410: \begin{example} \label{group}

411:   Let $G$ be a group, $X$ a set and $\theta$ an action of $G$ on $X$,

412:   i.e., $\theta_g$ is for every $g\in G$ a map from $X$ to $X$ such that

413:   \begin{subequations}    

414:     \begin{align}

415:       &\theta_e \text{ is the identity map on } X,\\

416:       &\theta_h\circ\theta_i=\theta_{hi} \text{ for every } h,i\in G.

417:     \end{align}

418:   \end{subequations}

419:   If we for every $g\in G$ let $D_g=X$, then $\theta=\bigl((D_g)_{g\in

420:     G},(\theta_g)_{g\in G}\bigr)$ is a partial action of $G$ on $X$, and

421:   $(X,\theta,G)$ a discrete partial dynamical system.

422: \end{example}

423: 

424: As mentioned in the Introduction, we will in this paper mainly concern

425: ourselves with partial dynamical systems which come from symbolic

426: dynamical systems. 

427: We will now show how to get partial dynamical systems from symbolic

428: dynamical systems; we will first see how to define a partial dynamical

429: system from a \emph{one-sided} symbolic dynamical system, and

430: then how to define a partial dynamical system from a \emph{two-sided}

431: symbolic dynamical system.  

432: 

433: \subsection{One-sided symbolic dynamical systems}\label{onesided}

434: Let $(\OSD,\osd)$ be a one-sided symbolic dynamical

435: system over the alphabet $\al$. That is: $\al$ is a set (finite or

436: infinite), $\osd:\al^{\No}\to \al^{\No}$ is the map 

437: \begin{equation*}

438:   x_0x_1x_2\dotsm\mapsto x_1x_2\dotsm

439: \end{equation*}

440: and $\OSD$ is a subset of $\al^{\No}$ such

441: that $\osd(\OSD)\subseteq \OSD$. To turn $(\OSD,\osd)$ into a

442: partial dynamical system we restrict $\osd$ to subsets of $\OSD$

443: such that $\osd$ is injective on these subsets. This is done in this way:

444: 

445: Let for every $a\in \al$, $D_a$ be the subset $\bigl\{(x_n)_{n\in\No}\in

446: \OSD\mid x_0=a\bigr\}$, $D_{a\inv}$ be the subset $\osd(D_a)$,

447: $\theta_a:D_{a\inv}\to D_a$ be the map 

448: \begin{equation*}

449:   x\mapsto ax,

450: \end{equation*}

451: and $\theta_{a\inv}:D_a\to D_{a\inv}$ be the map

452: \begin{equation*}

453:   x\mapsto \osd(x).

454: \end{equation*}

455: Let $\fg{\al}$ be the free group generated by $\al$ and let for every

456: $g\in \fg{\al}$ written in the reduced form $b_1b_2\dotsm b_k$, where

457: $b_1,b_2,\dotsc ,b_k\in \al\cup\al\inv$, $D_g$ be the subset of $\OSD$

458: defined by 

459: \begin{equation*}

460:   D_g=\theta_{b_1}\circ\theta_{b_2}\circ \dotsb \circ\theta_{b_k}(\OSD),

461: \end{equation*}

462: and $\theta_g:D_{g\inv}\to D_g$ be the map defined by

463: \begin{equation*}

464:   \theta_g=\theta_{b_1}\circ\theta_{b_2}\circ \dotsb \circ\theta_{b_k}.

465: \end{equation*}

466: Then $\theta_\OSD=\bigl((D_g)_{g\in\fg\al},(\theta_g)_{g\in\fg\al}\bigr)$ is a

467: partial action of $\fg\al$ on $\OSD$, and $(\OSD,\theta_\OSD,\fg{\al})$

468: is a discrete partial dynamical system.

469: 

470: \begin{definition} \label{onesideddef}

471:   Let $(\OSD,\osd)$ be a one-sided symbolic dynamical system over the

472:   alphabet $\al$. Then we

473:   call the discrete partial dynamical system

474:   $(\OSD,\theta_\OSD,\fg{\al})$ constructed above \emph{the

475:   discrete partial dynamical system associated to $(\OSD,\osd)$}.

476: \end{definition}

477: 

478: \begin{lemma} \label{lemma:onefund}

479:   Let $(\OSD,\osd)$ be a one-sided symbolic dynamical system over the

480:   alphabet $\al$ and let $(\OSD,\theta_\OSD,\fg{\al})$ be the discrete

481:   partial dynamical system associated to $(\OSD,\osd)$ as in

482:   Definition \ref{onesideddef}. Then the following holds for the

483:   partial action $\theta_\OSD=\bigl((D_g)_{g\in G},(\theta_g)_{g\in G}\bigr)$:

484:   \begin{enumerate}

485:   \item if $g\in G$ and $D_g\ne\emptyset$, then there exist

486:     $\mu,\nu\in\al^*$ such that $g=\mu\nu\inv$, \label{item:1}

487:   \item if $\mu,\nu\in\al^*$ and the last (rightmost) letters of $\mu$

488:     and $\nu$ are not equal (or either $\mu$ or $\nu$ is equal to the

489:     empty word), then we have that 

490:     \begin{equation*}

491:       D_{\nu\mu\inv}=\Bigl\{x\in\OSD\bigm| x_{[0,|\nu|[}=\nu,\ \mu

492:       x_{[|\nu|,\infty[}\in\OSD\Bigr\},

493:     \end{equation*}

494:     and that $\theta_{\nu\mu\inv}$ is the map 

495:     \begin{equation*}

496:       \mu x\mapsto \nu x

497:     \end{equation*}

498:     from $D_{\mu\nu\inv}$ to $D_{\nu\mu\inv}$, \label{item:2}

499:   \item if $\mu,\nu\in\al^*$, $|\mu|=|\nu|$ and $\mu\ne\nu$, then

500:     $D_\mu\cap D_\nu=\emptyset$. \label{item:3}

501:   \end{enumerate}

502: \end{lemma}

503: 

504: \begin{proof}

505:   (\ref{item:2}) can easily be proved by induction over the length of

506:   $\mu$ and $\nu$. (\ref{item:3}) then follows from (\ref{item:2}),

507:   and (\ref{item:1}) follows from the definition of $D_g$ and (\ref{item:3}).

508: \end{proof}

509: 

510: \subsection{Two-sided symbolic dynamical systems} \label{twosided}

511: Let $(\TSD,\tsd)$ be a two-sided symbolic dynamical system over the

512: alphabet $\al$. That is: $\al$ is a set (finite or infinite),

513: $\tsd:\al^{\Z}\to \al^{\Z}$ is the map defined by 

514: \begin{equation*}

515:   \bigl(\tsd((z_n)_{n\in\Z})\bigr)_m=z_{m+1}

516: \end{equation*}

517: for every $(z_n)_{n\in\Z}\in\al^{\Z}$ and every $m\in\Z$, and $\TSD$ is

518: a subset of $\al^{\Z}$ such that $\tsd(\TSD)=\TSD$. 

519: 

520: Since $\tsd$ is bijective, $(\tsd^k)_{k\in\Z}$ is an action of $\Z$ on

521: $\TSD$, so we could turn $(\TSD,\tsd)$ into a discrete partial

522: dynamical system by the method of Example \ref{group}, but we can also

523: do it by imitating the method we used to define the discrete partial

524: dynamical system associated to a one-sided symbolic dynamical system,

525: and that is what we will do here. 

526: 

527: Let for every $a\in \al$, $D_a$ be the subset

528: $\bigl\{(z_n)_{n\in\Z}\in \TSD\mid z_0=a\bigr\}$, $D_{a\inv}$ be the

529: subset $\bigl\{(z_n)_{n\in\Z}\in \TSD\mid z_{-1}=a\bigr\}$,

530: $\theta_a:D_{a\inv}\to D_a$ be the restriction of $\tsd\inv$ to

531: $D_{a\inv}$, and $\theta_{a\inv}:D_a\to D_{a\inv}$ be the restriction

532: of $\tsd$ to $D_a$. 

533:   

534: Let $\fg{\al}$ be the free group generated by $\al$, and let for every

535: $g\in \fg{\al}$ written in the reduced form $b_1b_2\dotsm b_k$, where

536: $b_1,b_2,\dotsc ,b_k\in \al\cup\al\inv$, $D_g$ be the subset of

537: $\TSD$ defined by

538: \begin{equation*}

539:   D_g=\theta_{b_1}\circ\theta_{b_2}\circ \dotsb \circ\theta_{b_n}(\TSD),

540: \end{equation*}

541: and let $\theta_g:D_{g\inv}\to D_g$ be the map defined by

542: \begin{equation*}

543:   \theta_g=\theta_{b_1}\circ\theta_{b_2}\circ \dotsb \circ\theta_{b_n}.

544: \end{equation*} 

545: Then $\theta_\TSD=\bigl((D_g)_{g\in\fg\al},(\theta_g)_{g\in\fg\al}\bigr)$

546: is a partial action of $\fg\al$ on $\TSD$, and $(\TSD,\theta_\TSD,\fg{\al})$

547: is a discrete partial dynamical system. 

548: 

549: \begin{definition} \label{twosideddef}

550:   Let $(\TSD,\tsd)$ be a two-sided symbolic dynamical system over the

551:   alphabet $\al$. Then we 

552:   call the discrete partial dynamical system

553:   $(\TSD,\theta_\TSD,\fg{\al})$ constructed above \emph{the

554:   discrete partial dynamical system associated to $(\TSD,\tsd)$}.

555: \end{definition}

556: 

557: \begin{lemma} \label{lemma:twofund}

558:   Let $(\TSD,\tsd)$ be a two-sided symbolic dynamical system over the

559:   alphabet $\al$ and let $(\TSD,\theta_\TSD,\fg{\al})$ be the discrete

560:   partial dynamical system associated to $(\TSD,\tsd)$ as in

561:   Definition \ref{twosideddef}. Then the following holds for the

562:   partial action $\theta_\TSD=\bigl((D_g)_{g\in G},(\theta_g)_{g\in G}\bigr)$:

563:   \begin{enumerate}

564:   \item if $g\in G$ and $D_g\ne\emptyset$, then either $g\in\al^*$ or

565:     $g\inv\in\al^*$, \label{item:t1}

566:   \item if $\mu\in\al^*$, then we have that

567:     \begin{equation*}

568:       D_\mu=\bigl\{z\in\TSD\mid z_{[0,|\mu|[}=u\bigr\},\quad

569:       D_{\mu\inv}=\bigl\{z\in\TSD\mid z_{[-|\mu|,0[}=u\bigr\}, 

570:     \end{equation*}

571:     $\theta_{\mu}$ is the restriction of $\tsd^{-|\mu|}$ to

572:     $D_{\mu\inv}$ and $\theta_{\mu\inv}$ is the restriction of

573:     $\tsd^{|\mu|}$ to $D_\mu$, \label{item:t2}

574:   \item if $\mu,\nu\in\al^*$, $|\mu|=|\nu|$ and $\mu\ne\nu$, then

575:     $D_\mu\cap D_\nu=\emptyset$ and $D_{\mu\inv}\cap

576:     D_{\nu\inv}=\emptyset$. \label{item:t3}

577:   \end{enumerate}

578: \end{lemma}

579: 

580: \begin{proof}

581:   (\ref{item:t2}) can easily be proved by induction over the length of

582:   $\mu$. (\ref{item:t3}) then follows from (\ref{item:t2}),

583:   and (\ref{item:t1}) follows from the definition of $D_g$ and (\ref{item:t3}).

584: \end{proof}

585: 

586: \section{The $\cs$-algebra associated to a discrete partial dynamical

587:   system}  \label{sec:cs-algebra}

588: The main object of this paper is to associate to every

589: discrete partial dynamical system $(X,\theta,G)$ a $\cs$-algebra

590: $\csp{X,\theta,G}$ in such a way that the class of

591: $\cs$-algebras obtained in this way in a natural way generalizes

592: Cuntz-Krieger algebras (both for finite and infinite matrices),

593: $\cs$-algebras associated to one-sided shift spaces and  

594: crossed products of two-sided shift spaces.

595: 

596: We want the $\cs$-algebra $\csp{X,\theta,G}$ associated to a discrete

597: partial dynamical system $(X,\theta,G)$ to have the

598: property that there is a bijective correspondence between the

599: representations of $\csp{X,\theta,G}$ and certain representations of

600: $(X,\theta,G)$. 

601: 

602: Let us as a motivating example look at the Cuntz-Krieger algebra $\O_A$ of an

603: $n\times n$-matrix $A=(A(i,j))_{i,j=1}^n$ with entries in $\{0,1\}$

604: and without any zero rows. The matrix $A$ gives raise to a one-sided

605: symbolic dynamical system, namely the topological Markov chain with

606: transition matrix $A$: 

607: \begin{equation*}

608:   \XA=\Big\{(x_n)_{n\in \No}\in\{1,2,,\dotsc,n\}^{\No}\bigm| \forall

609:   i\in\No:A(x_i,x_{i+1})=1\Big\}.

610: \end{equation*}

611: Thus we have by Definition \ref{onesideddef} a partial action $\theta_\XA$ of

612: the free group $\fg{n}$ generated by $n$ generators on $\XA$. It turns

613: out the structure of the discrete partial 

614: dynamical system $\xanp$ is reflected in the structure

615: of $\O_A$ in the following way: 

616: 

617: Let $(S_i)_{i=1}^n$ denote the generators of $\O_A$. Then the map 

618: \begin{equation*}

619:   i\mapsto S_i

620: \end{equation*}

621: extends to a partial representation $u$ of $\fg{n}$ on $\O_A$ (see Appendix

622: \ref{sec:parrep} for a short introduction to partial representations).

623: Since $u$ is a partial representation, $\bigl(u(g)\bigr)_{g\in\fg{n}}$ is a

624: family of partial isometries with commuting range projections, so

625: $\cs\bigl(\{u(g)u(g)^*\mid g\in \fg{n}\}\bigr)$ is a unital

626: abelian $\cs$-algebra, and so the set of projections in this

627: $\cs$-algebra is in a natural way a Boolean algebra (see Appendix

628: \ref{sec:boo} for a definition of this Boolean algebra). 

629: 

630: The discrete partial dynamical system $\xanp$ also gives raise to a Boolean

631: algebra: namely the Boolean algebra generated by 

632: all the domains $\{D_g\mid g\in \fg{n}\}$ of the partial action

633: $\theta_\XA$ (cf. Appendix \ref{sec:boo}). 

634: It turns out that the map

635: \begin{equation*}

636:   D_g\mapsto u(g)u(g)^*

637: \end{equation*}

638: extends to a Boolean homomorphism of the Boolean algebra generated by

639: the domains $\{D_g\mid g\in \fg{n}\}$ of the partial action $\theta_\XA$ to

640: the Boolean algebra of projections in the $\cs$-subalgebra of $\O_A$

641: generated by $\{u(g)u(g)^*\mid g\in \fg{n}\}$.

642: It is natural to view this Boolean homomorphism together with the

643: partial representation $u$ of $\fg{n}$ as a representation of the

644: partial dynamical system $\xanp$. We will in Section \ref{retu} see

645: that this representation of the partial dynamical system $\xanp$

646: on $\O_A$ is universal and thus completely characterizes $\O_A$.

647: 

648: Similar characterizations hold for Cuntz-Krieger algebras of infinite

649: matrices, the $\cs$-algebras associated to one-sided shift spaces and

650: the crossed products of two-sided shift spaces.

651: 

652: Thus we are lead to the following definition of

653: $\csp{X,\theta,G}$:

654: \begin{definition} \label{def:ox}

655:   Let $(X,\theta,G)$ be a discrete partial dynamical system. Then

656:   $\csp{X,\theta,G}$ is the universal $\cs$-algebra generated by

657:   a family $(s_g)_{g\in G}$ of elements satisfying:

658:   \begin{subequations}    

659:     \begin{align}

660:        &\parbox[t]{0.8\textwidth}{$(s_g)_{g\in

661:            G}$ is a family of partial isometries with commuting range

662:          projections,} \label{eq:ox}\\

663:       &s_e=1,\label{eq:oxa}\\

664:       &s_{g\inv}=s_g^* \text{ for every } g\in G,\label{eq:oxb}\\

665:       &s_hs_i=s_hs_h^*s_{hi} \text{ for every } h,i\in G,\label{eq:oxc}\\

666:       &\parbox[t]{0.8\textwidth}{the map $D_g\mapsto s_gs_g^*$ extends to a

667:         Boolean homomorphism from the Boolean algebra $\B$ generated

668:         by the domains $\{D_g\mid g\in G\}$ of the partial action

669:         $\theta$ to the Boolean algebra of projections in the unital

670:         abelian $\cs$-algebra $\cs\bigl(\{s_g

671:           s_g^*\mid g\in G\}\bigr)$.} \label{eq:oxd}   

672:     \end{align}

673:   \end{subequations}

674:   We will call the family $(s_g)_{g\in G}$ the \emph{generators} of

675:   $\csp{X,\theta,G}$. 

676: \end{definition} 

677: 

678: \begin{remark}

679: It is not immediately clear that a $\cs$-algebra with the

680: properties mentioned in Definition \ref{def:ox} exists,

681: but we will in Section \ref{sec:construction} for every  discrete

682: partial dynamical system constructed such a $\cs$-algebra as a

683: crossed product of a $\cs$-partial dynamical system.

684: \end{remark}

685: 

686: \begin{remark}

687:   It follows from Appendix \ref{sec:parrep} that condition

688:   \eqref{eq:ox}-\eqref{eq:oxc} are 

689:   equivalent to the following 3 conditions: 

690:   \begin{subequations}    

691:     \begin{align}

692:       &s_es_e^*=1,\\

693:       &s_g^*s_g=s_{g\inv}s_{g\inv}^* \text{ for every } g\in G,\\

694:       &s_hs_is_i^*s_h^*=s_hs_is_{hi}^* \text{ for every } h,i\in G

695:     \end{align}

696:   \end{subequations}

697:   (cf. \cite{MR99a:46121}) and with the following 4 conditions:

698:   \begin{subequations}    

699:     \begin{align}

700:       &\parbox[t]{0.8\textwidth}{$(s_g)_{g\in G}$

701:         is a family of partial isometries with commuting range

702:         projections,}\\

703:       &s_e=1,\\

704:       &s_{g\inv}=s_g^* \text{ for every }g\in G,\\

705:       &s_hs_is_{i\inv}=s_{hi}s_{i\inv} \text{ for every }h,i\in G,

706:     \end{align}

707:   \end{subequations}

708:   (cf. \cite{MR2003f:46108}).

709: \end{remark}

710: 

711: \begin{definition} \label{remarkboomap}

712:   Let $(X,\theta,G)$ be a discrete partial dynamical system, 

713:   $(D_g)_{g\in G}$ the domains of $\theta$ and $(s_g)_{g\in G}$ the

714:   generators of $\csp{X,\theta,G}$. It is clear that the Boolean

715:   homomorphism from $\B$ to the Boolean algebra of projections in

716:   $\cs\bigl(\{s_gs_g^*\in\csp{X,\theta,G}\mid g\in G\}\bigr)$ which

717:   extends the map $D_g\mapsto s_gs_g^*$ is unique. We will denote it

718:   by $\boomap$.  

719: \end{definition}

720: We will in Corollary \ref{boomapinj} see that $\boomap$ is always

721: injective, and thus that $\csp{X,\theta,G}\ne 0$ (unless $X=\emptyset$).

722: 

723: The following lemma describe how the structure of the discrete partial

724: dynamical system $(X,\theta,G)$ is reflected in the structure of the

725: $\cs$-algebra $\csp{X,\theta,G}$.

726: 

727: \begin{lemma} \label{booinvariant}

728:   If $(X,\theta,G)$ is a discrete partial dynamical system, then we

729:   have that 

730:   \begin{equation*}

731:      \boomap\bigl(\theta_g(A)\bigr)=s_g\boomap(A)s_g^*

732:   \end{equation*}

733:   for every $A\in\B$ and every $g\in G$.

734: \end{lemma}

735: 

736: \begin{proof}

737:   If $\boomap\bigl(\theta_g(A)\bigr)=s_g\boomap(A)s_g^*$ and

738:   $\boomap\bigl(\theta_g(B)\bigr)=s_g\boomap(B)s_g^*$, then the

739:   following series of equalities hold:

740:   \begin{equation*}

741:     \begin{split}

742:       \boomap\bigl(\theta_g(A\cap B)\bigr)&=

743:       \boomap\bigl(\theta_g(A)\cap\theta_g(B)\bigr)\\

744:       &= \boomap\bigl(\theta_g(A)\bigr)\boomap\bigl(\theta_g(B)\bigr)\\

745:       &= s_g\boomap(A)s_g^*s_g\boomap(B)s_g^*\\

746:       &= s_gs_g^*s_g\boomap(A)\boomap(B)s_g^*\\

747:       &= s_g\boomap\bigl(A\cap B\bigr)s_g^*,

748:     \end{split}

749:   \end{equation*}

750:   and so does the next:

751:   \begin{equation*}

752:     \begin{split}

753:       \boomap\bigl(\theta_g(X\setminus A)\bigr) &=

754:       \boomap\bigl(D_g\setminus\theta_g(A)\bigr)\\

755:       &= s_gs_g^*-s_g\boomap(A)s_g^*\\

756:       &= s_g\bigl(1-\boomap(A)\bigr)s_g^*\\

757:       &= s_g\boomap(X\setminus A)s_g^*.

758:     \end{split}

759:   \end{equation*}

760:   So it is enough to show that

761:   $\boomap\bigl(\theta_g(D_h)\bigr)=s_gs_hs_h^*s_g^*$ for every $g,h\in G$.

762: 

763:   So let $g,h\in G$. Then we have that

764:   \begin{equation*}

765:     \begin{split}

766:       \boomap\bigl(\theta_g(D_h)\bigr)

767:       &= \boomap\bigl(D_g\cap D_{gh}\bigr)\\

768:       &= s_gs_g^*s_{gh}s_{gh}^*\\

769:       &= s_gs_g^*s_{gh}s_{gh}^*s_gs_g^*\\

770:       &= s_gs_hs_h^*s_g^*,\\

771:     \end{split}

772:   \end{equation*}

773:   which proves the lemma.

774: \end{proof}

775: 

776: The next lemma, which we state for further reference, is an easy but

777: useful consequent of the definition of the $\cs$-algebra of a discrete

778: partial dynamical system.

779: 

780: \begin{lemma} \label{lemma:orto}

781:   Let $(X,\theta,G)$ be a discrete partial dynamical system and let

782:   $g,h,i\in G$. If $D_g=\emptyset$, then $s_g=0$, and if $D_h\cap

783:   D_i=\emptyset$, then $s_h^*s_i=0$.

784: \end{lemma}

785: 

786: \begin{proof}

787:   If $D_g=\emptyset$, then $s_g=s_gs_g^*s_g=\boomap\bigl(D_g\bigr)s_g=0$,

788:   and if $D_h\cap D_i=\emptyset$, then we have that

789:   \begin{equation*}

790:     \begin{split}

791:       s_h^*s_i&=s_h^*s_hs_h^*s_is_i^*s_i\\

792:       &=s_h^*\boomap\bigl(D_h\bigr)\boomap\bigl(D_i\bigr)s_i\\

793:       &=s_h^*\boomap\bigl(D_h\cap D_i\bigr)s_i\\

794:       &=0.

795:     \end{split}

796:   \end{equation*}

797: \end{proof}

798: 

799: \begin{example}

800:   If $\theta$ is an action of a discrete group $G$ on a set $X\ne

801:   \emptyset$ and 

802:   $(X,\theta,G)$ is the discrete partial dynamical system defined in

803:   Example \ref{group}, then all the domains $(D_g)_{g\in G}$ of the

804:   partial action $\theta$ are equal to $X$. Thus the Boolean algebra

805:   generated by these domains only consists of $X$ and $\emptyset$, and so

806:   $\cs(X,\theta,G)$ is just the group $\cs$-algebra of $G$. 

807: \end{example}

808: 

809: \section{A construction of $\csp{X,\theta,G}$}

810: \label{sec:construction} 

811: 

812: We will in this section for every discrete partial dynamical system

813: $(X,\theta,G)$ construct the $\cs$-algebra $\csp{X,\theta,G}$

814: (cf. Definition \ref{def:ox}) as a crossed product of a $\cs$-partial

815: dynamical systems (see Appendix \ref{sec:parcro} for an short

816: introduction to crossed product of $\cs$-partial dynamical systems,

817: cf. also \cites{MR2003f:46108,MR99a:46121,MR1331978}). 

818: 

819: We will first from the discrete partial dynamical system

820: $(X,\theta,G)$ construct the $\cs$-partial dynamical systems

821: $(\A,G,\ba{\theta})$ which $\csp{X,\theta,G}$ is a crossed product of. 

822: 

823: Let $(D_g)_{g\in G}$ denote the domains and $(\theta_g)_{g\in G}$ the

824: partial one-to-one maps of $\theta$. Recall that $\B$ is the Boolean

825: algebra generated by the domains $\{D_g\mid g\in G\}$ of the partial

826: action $\theta$. Notice that $\theta_h(D_i)=D_h\cap D_{hi}\in \B$ for

827: every $h,i\in G$, so $\theta_g(A)\in \B$ for every $A\in \B$ and every $g\in G$.

828:  

829: Let $\X$ be the dual of $\B$: i.e., $\X$ is the closed subset

830: \begin{equation*}

831:   \Bigl\{\phi\in\{0,1\}^{\B}\bigm| \phi \text{ is a Boolean homomorphism}\Bigr\}

832: \end{equation*}

833: of the Cantor space $\{0,1\}^{\B}$ endowed with the product topology

834: of the discrete topology on $\{0,1\}$. Then $\X$ is a totally

835: disconnected compact Hausdorff space (a \emph{Boolean space},

836: cf. \cite{MR29:4713}*{\S 18}). 

837: 

838: For each $A\in \B$, let $\h{A}=\{\phi\in \X\mid \phi(A)=1\}$, and

839: notice that $\h{A}$ is a clopen subset of $\X$. We then have that the

840: map $A\mapsto \h{A}$ is a Boolean isomorphism between $\B$ and the

841: Boolean algebra of clopen subsets of $\X$ (cf. \cite{MR29:4713}*{\S18}).

842: 

843: \begin{lemma} \label{separet}

844:   The system $\{\h{D}_t\mid t\in G\}$ separates points in $\X$.

845: \end{lemma}

846: 

847: \begin{proof}

848:   Let $\phi_1,\phi_2\in \X$ and let $\Aa$ be the subset of $\B$

849:   defined by

850:   \begin{equation*}

851:     \Aa=\{A\in \B\mid \phi_1(A)=\phi_2(A)\}.

852:   \end{equation*}

853:   Then $\Aa$ is a Boolean subalgebra of $\B$. Assume that the

854:   following equivalence 

855:   \begin{equation*}

856:     \phi_1\in\h{D}_g\iff \phi_2\in \h{D}_g

857:   \end{equation*}

858:   holds for every $g\in G$. That means that $D_g\in \Aa$ for every

859:   $g\in G$ and thus that $\Aa=\B$. But then $\phi_1$ and $\phi_2$ must

860:   be equal.  

861: \end{proof}

862: For each $g\in G$, let $\h{\theta}_g$ be the map given by

863: \begin{equation*}

864:   \h{\theta}_g(\phi)(A)=\phi\bigl(\theta_g\inv(A)\bigr)

865: \end{equation*}

866: for $A\in \B$ and $\phi\in \h{D}_{g\inv}$. It is easy to check that

867: $\h{\theta}_g$ is a homeomorphism from $\h{D}_{g\inv}$ to $\h{D}_{g}$

868: with $\h{\theta}_{g\inv}$ as its inverse (cf. \cite{MR29:4713}*{\S 20}). 

869: 

870: Let $\A=C(\h{X})$ and let for each $g\in G$, $\ba{D}_g$ be the subset

871: of $\A$ defined by

872: \begin{equation*}

873:   \ba{D}_g=\left\{f\in \A\bigm| f_{|\h{X}\setminus \h{D}_g}=0\right\}, 

874: \end{equation*}

875: and let $\ba{\theta}_g:\ba{D}_{g\inv}\to\ba{D}_g$ be defined by

876: \begin{equation*}

877:   \ba{\theta}_g(f)(\phi)=

878:   \begin{cases}

879:     f(\h{\theta}_{g\inv}(\phi))&\text{if }\phi\in\h{D}_g,\\

880:     0&\text{if }\phi\in \h{X}\setminus\h{D}_g,

881:   \end{cases}

882: \end{equation*}

883: for $f\in \ba{D}_{g\inv}$ and $\phi\in \h{X}$.

884: Then $\ba{\theta}=\bigl((\ba{D}_g)_{g\in G},(\ba{\theta}_g)_{g\in G}\bigr)$

885: is a partial action of $G$ on the $\cs$-algebra $\A$

886: (cf. Appendix \ref{sec:parcro}). Thus $(\A,G,\ba{\theta})$ is a $\cs$-partial

887: dynamical system. 

888: 

889: \begin{theorem} \label{hoved}

890:   Let $(X,\theta,G)$ be a discrete partial dynamical

891:   system and let $\A$ and $\ba{\theta}$ be as defined above. Then the

892:   partial crossed product $\A\rtimes_{\ba{\theta}}G$ of 

893:   the $\cs$-partial dynamical system $(\A,G,\ba{\theta})$ is

894:   generated by a family of elements $(s_g)_{g\in G}$ satisfying condition

895:   \eqref{eq:ox}--\eqref{eq:oxd} of Definition \ref{def:ox}, and if

896:   $\Aa$ is another $\cs$-algebra with a family of elements $(S_g)_{g\in G}$

897:   satisfying condition \eqref{eq:ox}--\eqref{eq:oxd}, then there

898:   exists a $*$-homomorphism from $\A\rtimes_{\ba{\theta}}G$ to $\Aa$

899:   which maps $s_g$ to $S_g$ for every $g\in G$.

900:  \end{theorem}

901: 

902: \begin{proof}

903:   Let for every $g\in G$, $s_g=\delta_g$, where $\{\delta_g\}_{g\in G}$

904:   is a described in Appendix \ref{sec:parcro}. Since $(\pi,u)\mapsto

905:   \pi\times u$ is a bijective correspondence between covariant

906:   representations of $(\A,G,\ba{\theta})$ and non-degenerated

907:   representations of $\A\rtimes_{\ba{\theta}}G$, there exists a

908:   covariant representation $(\pi,u)$ of $(\A,G,\ba{\theta})$ on a

909:   Hilbert space $\H$ such that $\pi\times u$ is a faithful

910:   non-degenerate representation of $\A\rtimes_{\ba{\theta}}G$. We then

911:   have that  $u$ is a partial representation of $G$ and so the family

912:   $\bigl(u(g)\bigr)_{g\in G}$ satisfies condition

913:   \eqref{eq:ox}--\eqref{eq:oxc} of Definition \ref{def:ox}.  

914: 

915:   Denote for every subset $B$ of $\A$, the subspace

916:   \begin{equation*}

917:     \spc\{\pi(T)\xi\mid T\in B,\ \xi\in \H\} 

918:   \end{equation*}

919:   of $\H$ by $[B]$ and the projection of $\H$ onto $[B]$ by $\proj([B])$. 

920: 

921:   Since the map sending an element $A$ of $\B$ to $\h{A}$, the map

922:   sending a clopen subset $V$ of $\h{X}$ to $\bigl\{f\in \A\bigm|

923:   f_{|\h{X}\setminus V}=0\bigr\}$, and the map sending an ideal $I$ of $\A$

924:   to $\proj([I])$ all are Boolean homomorphism, so is the map sending

925:   an element $A$ of $\B$ to

926:   $\proj\bigl(\bigl[\bigl\{f\in \A\mid f_{|\h{X}\setminus \h{A}}=

927:   0\bigr\}\bigr]\bigr)$, 

928:   and since we have that

929:   \begin{equation*}

930:     \begin{split}

931:       \proj\bigl(\bigl[\bigl\{f\in \A\mid f_{|\h{X}\setminus \h{D}_g}=

932:       0\bigr\}\bigr]\bigr)

933:       &=\proj([\ba{D}_g])\\

934:       &=u(g)u(g)^*

935:     \end{split}

936:   \end{equation*}

937:   for every $g\in G$, $\bigl(u(g)\bigr)_{g\in G}$ satisfies \eqref{eq:oxd} of

938:   Definition \ref{def:ox}. Since $\pi\times u$ is faithful and

939:   $\pi\times u(s_g)=u(g)$ for every $g\in G$, $(s_g)_{g\in G}$

940:   satisfies \eqref{eq:ox}--\eqref{eq:oxd} of Definition \ref{def:ox}. 

941: 

942:   Remember that $\A\rtimes_{\ba{\theta}}G$ is generated by $\A$ and

943:   $(s_g)_{g\in G}$. It follows from Lemma \ref{separet} and

944:   the Stone-Weierstrass Theorem that the equality 

945:   \begin{equation*}

946:     \spc\{1_{\h{D}_g}\mid g\in G\}=\A

947:   \end{equation*}

948:   holds, and since $s_gs_g^*=1_{\h{D}_g}$ for every $g\in G$, this

949:   shows that $\A\rtimes_{\ba{\theta}}G$ is generated by $(s_g)_{g\in G}$.  

950: 

951:   Now let $\Aa$ be another $\cs$-algebra with a family 

952:   $(S_g)_{g\in G}$ of elements which satisfies condition

953:   \eqref{eq:ox}--\eqref{eq:oxd}. Let $\psi$ be a non-degenerate

954:   faithful representation of $\Aa$ on a Hilbert space $\H$. Since

955:   $S_e$ is the unit of $\Aa$, $\psi(S_e)\xi=\xi$ for every $\xi\in\H$.

956: 

957:   Let for every $g\in G$, $U(g)=\psi(S_g)$. Then $U$ is a partial

958:   representation of $G$ on $\H$. Since $\spc\{1_{\h{D}_g}\mid g\in

959:   G\}=\A$, since the map $A\mapsto \h{A}$ is a Boolean isomorphism between

960:   $\B$ and the Boolean algebra of clopen subsets of $\X$, and since the map

961:   $D_g\mapsto S_gS_g^*$ extends to a Boolean homomorphism from $\B$ to

962:   the set of projections in $\cs\bigl(\{S_g S_g^*\mid g\in G\}\bigr)$,

963:   there exists by Lemma \ref{lemma:homo} a $*$-homomorphism

964:   from $\A$ to $\Aa$ which maps $1_{\h{D}_g}$ to $S_gS_g^*$ for every

965:   $g\in G$. Let us denote the composition of this $*$-homomorphism

966:   with $\psi$ by $\eta$. Then $\eta$ is a representation of $\A$ on

967:   $\H$, and since $\eta(1)\xi=\psi(S_e)\xi=\xi$ for every $\xi\in\H$,

968:   $\eta$ is non-degenerate.

969: 

970:   If $g\in G$, then we have that

971:   \begin{equation*}

972:     \begin{split}

973:       U(g)U(g)^*&=\psi(S_gS_g^*)\\

974:       &=\eta(1_{\h{D}_g})\\

975:       &=\proj\Bigl(\spc\bigl\{\eta(1_{\h{D}_g})\xi\mid \xi\in H\bigr\}\Bigr)\\

976:       &=\proj\bigl([\ba{D}_g]\bigr).    

977:     \end{split}

978:   \end{equation*}   

979:   Let $i,h\in G$. It then follows from \eqref{eq:paractb} that

980:   $\ba{\theta}_h\bigl(1_{\h{D}_{h\inv}\cap\h{D}_i}\bigr)=

981:   1_{\h{D}_h\cap\h{D}_{hi}}$. Thus we have that 

982:   \begin{equation*}

983:     \begin{split}

984:       \eta\Bigl(\ba{\theta}_h\bigl(1_{\h{D}_{h\inv}\cap\h{D}_i}\bigr)\Bigr)

985:       &=\eta\bigl(1_{\h{D}_h\cap\h{D}_{hi}}\bigr)\\ 

986:       &=U(h)U(h)^*U(hi)U(hi)^*\\

987:       &=U(h)U(i)U(hi)^*\\

988:       &=U(h)U(i)U(i\inv)U(i\inv)^*U(i\inv h\inv)\\

989:       &=U(h)U(h)^*U(h)U(i)U(i\inv)U(h\inv)\\

990:       &=U(h)\eta\bigl(1_{\h{D}_{h\inv}\cap\h{D}_i}\bigr)U(h\inv),

991:     \end{split}

992:   \end{equation*}

993:   where the third equality follows from \eqref{eq:oxc}, the forth

994:   from \eqref{eq:ox} and \eqref{eq:oxb}, and the fifth from

995:   \eqref{eq:ox} and \eqref{eq:oxd}. It follows from Lemma \ref{separet} and

996:   the Stone-Weierstrass Theorem that 

997:   $\spc\bigl\{1_{\h{D}_{h\inv}\cap\h{D}_i}\mid i\in

998:   G\bigr\}=\ba{D}_{h\inv}$, so the above computation shows that the

999:   equality 

1000:   \begin{equation*}

1001:     \eta\bigl(\ba{\theta}_h(f)\bigr)=U(h)\bigl(\eta(f)\bigr)U(h\inv)

1002:   \end{equation*}

1003:   holds for every $f\in \ba{D}_{h\inv}$. 

1004:   

1005:   Thus $(\eta,U)$ is a covariant representation of

1006:   $(\A,G,\ba{\theta})$, and so there exists a non-degenerate

1007:   representation of $\A\rtimes_{\ba{\theta}}G$ on $\H$ which maps

1008:   $s_g$ to $\psi(S_g)$ for every $g\in G$, and thus a

1009:   $*$-homomorphism from  $\A\rtimes_{\ba{\theta}}G$ to $\Aa$ which maps

1010:   $s_g$ to $S_g$ for every $g\in G$.

1011: \end{proof}

1012: 

1013: \begin{remark} \label{remark:exellaca}

1014:   Let $\ind$ be an arbitrary index set and let

1015:   $A={A(i,j)}_{i,j\in\ind}$ be a matrix with entries in $\{0,1\}$ and

1016:   having no zero rows. In \cite{MR2000i:46064} Exel and Lace

1017:   associated to $A$ a $\cs$-algebra $\O_A$ (cf. Section

1018:   \ref{subsec:CK}). We will in Theorem \ref{theorem:CK} show that the

1019:   unitization $\widetilde{\O}_A$ of $\O_A$ is isomorphic to

1020:   $\cspxa$ for a certain discrete partial dynamical system $\pxa$. 

1021: 

1022:   Exel and Laca constructed in \cite{MR2000i:46064} a partial action 

1023:   $\left((\Delta^A_g)_{g\in\fg\ind},(h^A_g)_{g\in\fg\ind}\right)$

1024:   of $\fg\ind$ on a compact topological space

1025:   $\widetilde{\Omega}_A$ such that $\widetilde{\O}_A$ is isomorphic to

1026:   the crossed product of this partial action. 

1027: 

1028:   If we let $(D_g)_{g\in G}$ and $(\theta_g)_{g\in G}$ denote the domains

1029:   and partial one-to-one maps of $\theta_\XA$, and $\h{\mathsf{X}}_A^+$,

1030:   $\h{D}_g$ and $\h\theta_g$ be as above, then one can show that the map

1031:   \begin{equation*}

1032:     \phi\mapsto \bigl\{g\in\fg\ind\mid \phi(D_g)=1\bigr\}

1033:   \end{equation*}

1034:   is a homeomorphism from $\h{\mathsf{X}}_A^+$ to

1035:   $\widetilde{\Omega}_A$ which for every 

1036:   $g\in\fg\ind$ maps $\h{D}_g$ to $\Delta^A_g$ and intertwines

1037:   $\h\theta_g$ and $h^A_g$. This fact lays the foundation for an

1038:   alternative proof of Theorem \ref{theorem:CK}.

1039: \end{remark}

1040: 

1041: \section{A representation of $\csp{X,\theta,G}$} \label{sec:repr}

1042: Let $(X,\theta,G)$ be a discrete partial dynamical system and let

1043: $(D_g)_{g\in G}$ denote the domains of $\theta$ and $(s_g)_{g\in G}$ the

1044: generators of $\csp{X,\theta,G}$. If $\pi$ is a representation of

1045: $\csp{X,\theta,G}$, then $\pi$ will induce a Boolean homomorphism from

1046: the Boolean algebra of projections in the unital abelian $\cs$-algebra

1047: $\cs\bigl(\{s_gs_g^*\in\csp{X,\theta,G}\mid g\in G\}\bigr)$ to the

1048: Boolean algebra of projections in the unital abelian $\cs$-algebra 

1049: $\cs\bigl(\{\pi(s_gs_g^*)\mid g\in G\}\bigr)$ which maps $s_gs_g^*$ to

1050: $\pi(s_gs_g^*)$ for every $g\in G$, and by composing this Boolean

1051: homomorphism with $\boomap$, we get a Boolean homomorphism from

1052: $\Bx(X,\theta,G)$ to the Boolean algebra of projections in the unital

1053: abelian $\cs$-algebra $\cs\bigl(\{\pi(s_gs_g^*)\mid g\in G\}\bigr)$

1054: which maps $D_g$ to $\pi(s_gs_g^*)$ for every $g\in G$. 

1055: 

1056: We will in this section for every discrete partial dynamical system 

1057: $(X,\theta,G)$ construct a representation $\pi$ of $\csp{X,\theta,G}$

1058: such that the Boolean homomorphism mentioned above is injective. As

1059: a corollary of this, we see that the Boolean homomorphism $\boomap$ is

1060: injective and thus that $\csp{X,\theta,G}\ne 0$ (unless $X=\emptyset$).

1061: 

1062: Let $(X,\theta,G)$ be a discrete partial dynamical system and let

1063: $(D_g)_{g\in G}$ and $(\theta_g)_{g\in G}$ denote the domains and partial

1064: one-to-one maps of $\theta$.

1065: Let $(e_x)_{x\in X}$ be an orthonormal basis for the Hilbert space

1066: $l_2(X)$, and define for each $g\in G$ an operator $S_g$ by letting 

1067: \begin{equation} \label{eq:op}

1068:   S_g\left(\sum_{x\in X}\lambda_xe_x\right)=\sum_{x\in

1069:     D_g}\lambda_{\theta_{g\inv}(x)}e_x 

1070: \end{equation}

1071: for every $\sum_{x\in X}\lambda_xe_x\in l_2(X)$.

1072: 

1073: It is straight forward to check that these operators are partial

1074: isometries, that $S_e=1$ and that for every $h,i\in G$, $S_h^*=S_{h\inv}$,

1075: $S_hS_i=S_hS_h^*S_{hi}$ and $S_hS_h^*=\proj\bigl(\spc\{e_x\mid x\in

1076: D_h\}\bigr)$, where $\proj\bigl(\spc\{e_x\mid x\in D_h\}\bigr)$ is the

1077: orthogonal 

1078: projection of $l_2(X)$ onto $\spc\{e_x\mid x\in D_h\}$. Since the map

1079: \begin{equation*}

1080:   A \mapsto \proj\bigl(\spc\{e_x\mid x\in A\}\bigr)

1081: \end{equation*}

1082: is a Boolean homomorphism from $\B$ to the Boolean algebra of

1083: projections in the unital abelian $\cs$-algebra 

1084: $\cs\bigl(\{S_gS_g^*\mid g\in G\}\bigr)$, the family $(S_g)_{g\in G}$

1085: of operators satisfies condition \eqref{eq:ox}--\eqref{eq:oxd} of

1086: Definition \ref{def:ox}. 

1087: 

1088: Thus we have:

1089: \begin{proposition}

1090:   Let $(X,\theta,G)$ be a discrete partial dynamical

1091:   system, let $(s_g)_{g\in G}$ denote the generators of $\csp{X,\theta,G}$

1092:   and $(S_g)_{g\in G}$ the operators defined by \eqref{eq:op}. Then

1093:   there is a $*$-homomorphism from $\csp{X,\theta,G}$ to 

1094:   the $\cs$-algebra of bounded operators on the Hilbert space

1095:   $l_2(X)$, sending $s_g$ to $S_g$ for every $g\in G$.

1096: \end{proposition}

1097: 

1098: \begin{corollary} \label{boomapinj}

1099:   The Boolean homomorphism $\boomap$ (cf. Definition \ref{remarkboomap})

1100:   is for every discrete partial dynamical system $(X,\theta,G)$ injective.

1101: \end{corollary}

1102: 

1103: \begin{proof}

1104:   Since the Boolean homomorphism

1105:   \begin{equation*}

1106:     A \mapsto \proj\bigl(\spc\{e_x\mid x\in A\}\bigr)

1107:   \end{equation*}

1108:   from $\B$ to the set of projections in

1109:   $\cs\bigl(\{S_g S_g^*\mid g\in G\}\bigr)$ is injective, so is $\boomap$.

1110: \end{proof}

1111: 

1112: \section{$\cs$-algebras associated to symbolic dynamical

1113:   systems} \label{retu} 

1114: We will now show that the class of $\cs$-algebras associated to discrete

1115: partial dynamical systems generalizes Cuntz-Krieger algebras (both

1116: for finite and infinite matrices), $\cs$-algebras associated to

1117: one-sided shift spaces and crossed products of two-sided shift spaces.

1118: 

1119: We will do that by regarding one-sided and two-sided shift spaces and

1120: topological Markov chains as symbolic dynamical systems and thus

1121: associated to them the discrete partial dynamical system of Definition

1122: \ref{onesideddef} and \ref{twosideddef} and then show that the

1123: $\cs$-algebras of these discrete partial dynamical systems are

1124: isomorphic to the $\cs$-algebra associated to the one-sided shift

1125: space,the crossed product of the two-sided shift space, and the

1126: unitization of the Cuntz-Krieger algebra of the transition matrix of

1127: the Markov chain, respectively.

1128: 

1129: Before we do that let us briefly look at the general structure of

1130: $\csp{X,\theta,G}$ when $(X,\theta,G)$ is the discrete partial

1131: dynamical system associated to a one- or two-sided symbolic dynamical

1132: system as in Definition \ref{onesideddef} and \ref{twosideddef}.

1133: 

1134: \begin{lemma} \label{lemma:onestruc}

1135:   Let $(\OSD,\osd)$ be a one-sided symbolic dynamical system over the

1136:   alphabet $\al$ and let $(\OSD,\theta_\OSD,\fg{\al})$ be the discrete

1137:   partial dynamical system associated to $(\OSD,\osd)$ as in

1138:   Definition \ref{onesideddef}. Then the following holds for the

1139:   generators $(s_g)_{g\in\fg\al}$ of $\csp{\OSD,\theta_\OSD,\fg{\al}}$:

1140:   \begin{enumerate}

1141:   \item $s_{b_1}s_{b_2}\dotsm s_{b_k}=s_g$ for $g\in

1142:     \fg\al$ written in reduced form, $b_1b_2\dotsm b_k$, \label{enum:oneet}

1143:   \item if $s_g\ne 0$, then there exist $\mu,\nu\in\al^*$ such that

1144:     $g=\mu\nu\inv$, \label{enum:oneto}

1145:   \item $s_\mu^*s_\nu=0$ if $\mu,\nu\in\al^*$, $|\mu|=|\nu|$ and

1146:     $\mu\ne\nu$. \label{enum:onetre}

1147:   \end{enumerate}

1148: \end{lemma}

1149: 

1150: \begin{proof}

1151:   If $b\in \al\cup\al\inv$ and $g\in \fg\al$ written in reduced form

1152:   does not begin with $b$, then $D_{bg}\subseteq D_b$, so we have that

1153:   \begin{equation*}

1154:     \begin{split}

1155:       s_bs_g&=s_bs_b^*s_{bg}\\

1156:       &= s_bs_b^*s_{bg}s_{bg}^*s_{bg}\\

1157:       &=s_{bg}.

1158:     \end{split}

1159:   \end{equation*}

1160:   This shows that $s_{b_1}s_{b_2}\dotsm s_{b_k}=s_g$ for

1161:   $g\in \fg\al$ written in reduced form $b_1b_2\dotsm b_k$.

1162: 

1163:    (\ref{enum:oneto}) and (\ref{enum:onetre}) easily follow from Lemma

1164:    \ref{lemma:onefund} and \ref{lemma:orto}.

1165: \end{proof}

1166: 

1167: \begin{lemma} \label{lemma:twostruc}

1168:   Let $(\TSD,\tsd)$ be a two-sided symbolic dynamical system over the

1169:   alphabet $\al$ and let $(\TSD,\theta_\TSD,\fg{\al})$ be the discrete

1170:   partial dynamical system associated to $(\TSD,\tsd)$ as in

1171:   Definition \ref{twosideddef}. Then the following holds for the

1172:   generators $(s_g)_{g\in\fg\al}$ of $\csp{\TSD,\theta_\TSD,\fg{\al}}$:

1173:   \begin{enumerate}

1174:   \item $s_{b_1}s_{b_2}\dotsm s_{b_k}=s_g$ for $g\in

1175:     \fg\al$ written in the reduced form $b_1b_2\dotsm b_k$, \label{enum:twoet}

1176:   \item if $s_g\ne 0$, then either $g\in\al^*$ or

1177:     $g\inv\in\al^*$, \label{enum:twoto}

1178:   \item $s_h^*s_i=s_hs_i^*=0$ if $h,i\in\fg\al$, $[h]=[i]$ and $h\ne

1179:     i$. \label{enum:twotre}

1180:   \end{enumerate}

1181: \end{lemma}

1182: 

1183: \begin{proof}

1184:   (\ref{enum:twoet}) can be proved in exactly

1185:   the same way (\ref{enum:oneet}) was proved

1186:   in Lemma \ref{lemma:onestruc}, (\ref{enum:twoto}) easily follows

1187:   from Lemma \ref{lemma:twofund} and \ref{lemma:orto}, and

1188:   (\ref{enum:twotre}) follows from (\ref{enum:twoto}) and Lemma

1189:   \ref{lemma:twofund} and \ref{lemma:orto}. 

1190: \end{proof}

1191: 

1192: \subsection{Crossed products of two-sided shift

1193:   spaces} \label{sec:cross} 

1194: Let $(\TSS,\tsh)$ be a two-sided shift space over the finite alphabet

1195: $\al$ (cf. \cite{MR1484730} and \cite{MR97a:58050}). That is:

1196: $\tsh:\al^{\Z}\to\al^{\Z}$ is the map defined by

1197: \begin{equation} \label{eq:twoshift}

1198:   \bigl(\tsh((z_n)_{n\in\Z})\bigr)_m=z_{m+1}

1199: \end{equation}

1200: for every $(z_n)_{n\in\Z}\in\al^{\Z}$ and every $m\in\Z$, and $\TSS$

1201: is a closed (in the product topology of the discrete topology of

1202: $\al$) subset of $\al^{\Z}$ such that $\tsh(\TSS)=\TSS$.

1203: 

1204: Let $\tsh^\star$ be the automorphism on $C(\TSS)$ defined by $f\mapsto

1205: f\circ\tsh$ and let $C(\TSS)\rtimes_{\tsh^\star}\Z$ be the full

1206: crossed product of the $\cs$-dynamical system $C(\TSS,\tsh^\star,\Z)$

1207: (cf. \cite{MR81e:46037}*{7.6.5}). Thus $C(\TSS)\rtimes_{\tsh^\star}\Z$

1208: is the universal $\cs$-algebra generated by a copy of $C(\TSS)$ and an

1209: unitary operator $U$ which satisfies that $UfU^*=f\circ \tsh$ for

1210: every $f\in C(\TSS)$.

1211: 

1212: Since $(\TSS,\tsh)$ is a two-sided symbolic dynamical system, we can

1213: associate to it the discrete partial dynamical system

1214: $(\TSS,\theta_\TSS,\fg{\al})$ of Definition \ref{twosideddef}. We then

1215: have the following theorem:

1216:  

1217: \begin{theorem} \label{kryds}

1218:   Let $(\TSS,\tsh)$ be a two-sided shift space and let

1219:   $(\TSS,\theta_\TSS,\fg{\al})$ be the discrete partial dynamical system

1220:   associated to $(\TSS,\tsh)$ as done in Definition \ref{twosideddef}. Then

1221:   $\csp{\TSS,\theta_\TSS,\fg{\al}}$ is isomorphic to the crossed product

1222:   $C(\TSS)\rtimes_{\tsh^\star}\Z$. 

1223: 

1224:   More precisely: if $(D_g)_{g\in\fg\al}$ denotes the domains of

1225:   $\theta_\TSS$, $(s_g)_{g\in\fg\al}$ denotes the generators of

1226:   $\csp{\TSS,\theta_\TSS,\fg{\al}}$ and $U$ is as above, then we have that

1227:   \begin{equation*}

1228:     C(\TSS)=\spc\{1_{D_g}\mid g\in \fg{\al}\},

1229:   \end{equation*} 

1230:   and there exists a $*$-isomorphism from $C(\TSS)\rtimes_{\tsh^\star}\Z$ to

1231:   $\csp{\TSS,\theta_\TSS,\fg{\al}}$ which maps $1_{D_g}$ to $s_gs_g^*$ for

1232:   every $g\in\fg\al$ and $U$ to $\sum_{a\in\al}s_a$.

1233: \end{theorem}

1234: 

1235: \begin{proof}

1236:   We will first show that the Boolean algebra

1237:   $\Bx(\TSS,\theta_\TSS,\fg\al)$ generated by 

1238:   $\{D_g\mid g\in \fg{\al}\}$ is the Boolean algebra of clopen subsets 

1239:   of $\TSS$. It easily follows from Lemma \ref{lemma:twofund} that 

1240:   every set in $\Bx(\TSS,\theta_\TSS,\fg\al)$ is clopen.  

1241:   In the other direction, we have by Lemma \ref{lemma:twofund} that

1242:   \begin{equation*}

1243:     \bigl\{z\in \TSS\mid z_{[-|u|,|v|[}=uv\bigr\}= D_{u\inv}\cap D_{v} \in

1244:     \Bx(\TSS,\theta_\TSS,\fg\al) 

1245:   \end{equation*}

1246:   for $u,v\in\al^*$, and since  the system

1247:   consisting of sets of this form is a basis for the topology of

1248:   $\TSS$, every clopen set is a finite union of sets of this form

1249:   and thus belongs to $\Bx(\TSS,\theta_\TSS,\fg\al)$.

1250:   

1251:   So it follows from the Stone-Weierstrass theorem that

1252:   \begin{equation*}

1253:     C(\TSS)=\spc\{1_{D_g}\mid g\in \fg{\al}\}. 

1254:   \end{equation*}

1255:   Since $\boomap[(\TSS,\theta_\TSS,\fg\al)]$ (cf. Definition

1256:   \ref{remarkboomap}) 

1257:   is a Boolean homomorphism from $\Bx(\TSS,\theta,\fg\al)$ to the Boolean

1258:   algebra of projections in the unital abelian $\cs$-algebra

1259:   \begin{equation*}

1260:     \cs\bigl(\{s_gs_g^*\in\csp{\TSS,\theta_\TSS,\fg\al}\mid g\in \fg\al\}\bigr) 

1261:   \end{equation*}

1262:   which maps $D_g$ to $s_gs_g^*$, it follows from Lemma \ref{lemma:homo}

1263:   that there exists a $*$-homomorphism $\crosmap$ from $C(\TSS)$ to

1264:   $\csp{\TSS,\theta_\TSS,\fg{\al}}$ which maps $1_{D_g}$ to $s_gs_g^*$

1265:   for every $g\in \fg\al$. Let $u$ be the element $\sum_{a\in\al}s_a$ in

1266:   $\csp{\TSS,\theta_\TSS,\fg{\al}}$. Then we have that

1267:   \begin{equation*}

1268:     \begin{split}

1269:       u\crosmap(1_{D_g})u^*& =\left(\sum_{a\in\al}s_a\right)

1270:       \bigl(s_gs_g^*\bigr) 

1271:       \left(\sum_{a'\in\al}s_{a'}^*\right) \\

1272:       &=\sum_{\substack{a\in\al\\

1273:           a'\in\al}}s_as_gs_g^*s_a^*s_as_{a'}^*s_{a'}s_{a'}^* \\      

1274:       &=\sum_{a\in\al}s_as_gs_g^*s_a^*\\

1275:       &=\sum_{a\in\al}s_as_a^*s_{ag}s_{ag}^*s_{a}s_{a}^*\\

1276:       &=\sum_{a\in\al}\crosmap\bigl(1_{D_a\cap D_{ag}}\bigr)\\

1277:       &=\sum_{a\in\al}\crosmap\bigl(1_{\theta_a(D_g)}\bigr)\\

1278:       &=\crosmap\bigl(1_{\cup_{a\in\al}\theta_a(D_g)}\bigr)\\

1279:       &=\crosmap\bigl(1_{\tsh\inv(D_g)}\bigr)\\

1280:       &=\crosmap(1_{D_g}\circ\tsh)

1281:     \end{split}

1282:   \end{equation*}

1283:   for every $g\in\fg\al$, where the second equality follows from

1284:   \eqref{eq:ox} and \eqref{eq:oxd}, the third from \eqref{eq:ox} and

1285:   Lemma \ref{lemma:twostruc}\eqref{enum:twotre}, the forth from

1286:   \eqref{eq:oxc} and the sixth from \eqref{eq:paractb}. Since

1287:   $C(\TSS)=\spc\{1_{D_g}\mid g\in \fg{\al}\}$, this shows that

1288:   \begin{equation*}

1289:     u\crosmap(f)u^*=\crosmap(f\circ\tsh)

1290:   \end{equation*} 

1291:   for every $f\in C(\TSS)$. 

1292:   

1293:   Thus it follows from the universal property

1294:   of $C(\TSS)\rtimes_{\tsh^\star}\Z$ that there exists a $*$-homomorphism

1295:   $\tcrosmap$ from $C(\TSS)\rtimes_{\tsh^\star}\Z$ to

1296:   $\csp{\TSS,\theta_\TSS,\fg{\al}}$ which is equal to $\crosmap$ on $C(\TSS)$

1297:   and sends $U$ to $u=\sum_{a\in\al}s_a$.

1298:   

1299:   Let us now look at $C(\TSS)\rtimes_{\tsh^\star}\Z$. Let for every

1300:   $a\in\al$, $S_a$ be the element $1_{D_a}U$ of

1301:   $C(\TSS)\rtimes_{\tsh^\star}\Z$, 

1302:   $S_{a\inv}=S_a^*$, and let $S_e=1$ and $S_g=S_{b_1}S_{b_2}\dotsm

1303:   S_{b_k}$, where $g\in \fg\al$ is written in the reduced form $b_1b_2\dotsm 

1304:   b_k$. Then $(S_g)_{g\in\fg{\al}}$ is a family of elements from

1305:   $C(\TSS)\rtimes_{\tsh^\star}\Z$ 

1306:   which clearly satisfies \eqref{eq:oxa} and \eqref{eq:oxb} 

1307:   of Definition \ref{def:ox}. We will now show that $(S_g)_{g\in

1308:     \fg{\al}}$ also satisfies \eqref{eq:ox}, \eqref{eq:oxc} and

1309:   \eqref{eq:oxd} of Definition \ref{def:ox}.

1310:   

1311:   Remember (cf. Section \ref{sec:notat}) that $[\cdot]$ is the unique

1312:   homomorphism from $\fg\al$ to $\Z$ such that $[a]=1$ for every $a\in\al$.

1313:   According to Lemma \ref{lemma:twofund}, the map $\theta_g$ is for every

1314:   $g\in\fg\al$ equal to 

1315:   the restriction of $\tsh^{-[g]}$ to $D_{g\inv}$, so we have that

1316:   \begin{equation*}

1317:     D_h\cap\tsh^{-[h]}(D_i)=\theta_h(D_i)=D_{hi}\cap D_h

1318:   \end{equation*}

1319:   for all $h,i\in\fg\al$.

1320:   

1321:   If $a\in\al$ and $g\in\fg\al$ written in reduced form does

1322:   not begin with $a\inv$, then $D_{ag}\subseteq D_a$ and

1323:   $D_a\cap\tsh^{\inv}(D_g)=D_{ag}\cap D_a=D_{ag}$ and so we have that

1324:   \begin{equation*}

1325:     \begin{split}

1326:       S_a1_{D_g}U^{[g]}&=1_{D_a}U1_{D_g}U^{[g]}\\

1327:       &=1_{D_a}U1_{D_g}U^*U^{[g]+1}\\

1328:       &=1_{D_a\cap\tsh\inv(D_g)}U^{[g]+1}\\

1329:       &=1_{D_{ag}}U^{[g]+1},\\

1330:     \end{split}

1331:   \end{equation*}

1332:   and if $g$ written in reduced form does not begin with an $a$,

1333:   then $D_{a\inv g}\subseteq D_{a\inv}$ and $\tsh(D_a\cap

1334:   D_g)=\theta_{a\inv}(D_g)=D_{a\inv}\cap D_{a\inv g}=D_{a\inv g}$,

1335:   so we have that

1336:   \begin{equation*}

1337:     \begin{split}

1338:       S_{a\inv}1_{D_g}U^{[g]}&=U^*1_{D_a}1_{D_g}U^{[g]}\\

1339:       &=1_{\tsh(D_a\cap D_g)}U^{[g]-1}\\

1340:       &=1_{D_{a\inv g}}U^{[g]-1}.

1341:     \end{split}

1342:   \end{equation*}

1343:   This shows that $S_g=1_{D_g}U^{[g]}$ and thus that $S_gS_g^*=1_{D_g}$

1344:   for every $g\in \fg\al$. Hence $(S_g)_{g\in\fg{\al}}$ satisfies

1345:   \eqref{eq:ox} and \eqref{eq:oxd}  

1346:   of Definition \ref{def:ox}. If $h,i\in\fg\al$, then we have that

1347:   \begin{equation*}

1348:     \begin{split}

1349:       S_hS_i&=1_{D_h}U^{[h]}1_{D_i}U^{[i]}\\

1350:       &=1_{D_h}U^{[h]}1_{D_i}U^{-[h]}U^{[h]}U^{[i]}\\

1351:       &=1_{D_h\cap\tsh^{-[h]}(D_i)}U^{[hi]}\\

1352:       &=1_{D_{hi}\cap D_h}U^{[hi]}\\

1353:       &=1_{D_h}1_{D_{hi}}U^{[hi]}\\

1354:       &=S_hS_h^*S_{hi},

1355:     \end{split}

1356:   \end{equation*}

1357:   which shows that $(S_g)_{g\in\fg{\al}}$ also satisfies

1358:   \eqref{eq:oxc} of Definition \ref{def:ox}.

1359:   

1360:   Thus it follows from the universal property of

1361:   $\csp{\TSS,\theta_\TSS,\fg{\al}}$ that there is a

1362:   $*$-homomorphism $\psi$ from $\csp{\TSS,\theta_\TSS,\fg{\al}}$ to

1363:   $C(\TSS)\rtimes_{\tsh^\star}\Z$ such that

1364:   $\psi(s_g)=S_g=1_{D_g}U^{[g]}$ for every $g\in \fg\al$.

1365:   

1366:   We have that 

1367:   \begin{equation*}

1368:     \psi\bigl(\tcrosmap(U)\bigr)= \psi\bigg(\sum_{a\in\al}s_a\bigg)=

1369:     \sum_{a\in\al}1_{D_a}U=U,

1370:   \end{equation*}

1371:   and that

1372:   \begin{equation*}

1373:     \psi\bigl(\tcrosmap(1_{D_g})\bigr)=

1374:     \psi(s_gs_g^*)= 1_{D_g}U^{[g]}U^{-[g]}1_{D_g}= 1_{D_g}

1375:   \end{equation*}

1376:   for every $g\in \fg\al$, and since $C(\TSS)\rtimes_{\tsh^\star}\Z$

1377:   is generated by $U$ and $\{1_{D_g}\mid g\in\fg{\al}\}$, this shows that

1378:   $\psi\circ\tcrosmap=\Id_{C(\TSS)\rtimes_{\tsh^\star}\Z}$.

1379:   

1380:   We also have that

1381:   \begin{equation*}

1382:     \tcrosmap\bigl(\psi(s_g)\bigr)=\tcrosmap\bigl(1_{D_g}U^{[g]}\bigr)=

1383:     s_gs_g^*\bigl(\sum_{a\in\al}s_a\bigr)^{[g]}

1384:   \end{equation*} 

1385:   for every $g\in \fg\al$. It follows from Lemma

1386:   \ref{lemma:twostruc} that we have that

1387:   \begin{equation*}

1388:     \biggl(\sum_{a\in\al}s_a\biggr)^{[g]}=\sum_{\substack{h\in\fg\al\\ [h]=[g]}}s_h,

1389:   \end{equation*}

1390:   and that 

1391:   \begin{equation*}

1392:     s_gs_g^*s_h=

1393:     \begin{cases}

1394:       s_g&\text{if }g=h,\\

1395:       0&\text{if }g\ne h,

1396:     \end{cases}

1397:   \end{equation*}

1398:   for every $h\in \fg\al$ with $[h]=[g]$. Thus it follows that

1399:   \begin{equation*}

1400:     \tcrosmap(\psi(s_g))=s_gs_g^*\biggl(\sum_{a\in\al}s_a\biggr)^{[g]}=s_g

1401:   \end{equation*}

1402:   for every $g\in\fg\al$, which shows that

1403:   $\tilde{\crosmap}\circ\psi=\Id_{\csp{\TSS,\theta_\TSS,\fg{\al}}}$. 

1404:   

1405:   Hence $\tcrosmap$ is an isomorphism from

1406:   $C(\TSS)\rtimes_{\tsh^\star}\Z$ to $\csp{\TSS,\theta_\TSS,\fg{\al}}$

1407:   which maps $1_{D_g}$ to $s_gs_g^*$ for

1408:   every $g\in\fg\al$ and $U$ to $\sum_{a\in\al}s_a$.

1409: \end{proof}

1410: 

1411: \subsection{$\cs$-algebras associated to one-sided shift

1412:   spaces} \label{sec:onesided} 

1413: Let $(\OSS,\osh)$ be a one-sided shift space over the finite alphabet

1414: $\al$ (cf. \cite{MR1484730} and \cite{MR97a:58050}*{\S13.8}). That is:

1415: $\osh:\al^{\No}\to\al^{\No}$ is the map

1416: \begin{equation} \label{eq:oneshift}

1417:   x_0x_1x_2\dotsm\mapsto x_1x_2\dotsm,

1418: \end{equation}

1419: and $\OSS$ is a closed (in the product topology of the discrete

1420: topology of $\al$) subset of $\al^{\No}$ such that $\tsh(\OSS)=\OSS$.

1421:   

1422: As far as the author know, Kengo Matsumoto was in

1423: \cite{MR1454478} the first to consider $\cs$-algebras associated

1424: to shift spaces. Matsumoto's construction is however in the opinion of

1425: the author not the optimal one (see \cite{MR2091486} for a discussing

1426: of this matter). In \cite{tmc} the author considered a different

1427: construction of $\cs$-algebras associated to shift spaces which for

1428: some shift spaces gives a slightly different $\cs$-algebra than

1429: Matsumoto's (and the $\cs$-algebra considered in \cite{MR2091486},

1430: cf. \cite{CS}*{Section 7}). We will in this paper work with the

1431: $\cs$-algebra $\O_{\OSS}$  of \cite{tmc} (it is isomorphic to the $\cs$-algebra

1432: $\Dx\negthickspace\rtimes_{\alpha,\mathcal{L}}\N$ considered in \cite{CS}, cf.

1433: \cite{CS}*{Remark 9}). It can be characterized in the following way

1434: (cf. \cite{tmc}*{Remark 7.3}):

1435: 

1436: Let $\al^*$ denote the set of finite words with letters from $\al$.

1437: For $\mu,\nu\in \al^*$, let $\cy\mu\nu$ be the subset of $\OSS$

1438: defined by 

1439: \begin{equation*}

1440:   \cy\mu\nu=\bigl\{x\in \OSS\bigm| x_{[0,|\nu|[}=\nu,\ \mu

1441:   x_{[|\nu|,\infty[}\in \OSS\bigr\}.

1442: \end{equation*}

1443: We let $\mathfrak{B}(\OSS)$ be the abelian

1444: $C^*$-algebra of all bounded functions on $\OSS$, and $\Dx$ the

1445: $C^*$-subalgebra of $\mathfrak{B}(\OSS)$ generated by

1446: $\bigl\{1_{\cy\mu\nu}\mid \mu,\nu \in\al^*\bigr\}$.

1447: 

1448: Then the $\cs$-algebra $\O_{\OSS}$ associated to the one-sided shift

1449: space $(\OSS,\osh)$ is the universal $\cs$-algebra generated by a

1450: family of partial isometries $(S_a)_{a\in\al}$ which satisfies that the map

1451: \begin{equation*}

1452:   1_{\cy\mu\nu}\mapsto S_\nu S_\mu^*S_\mu S_\nu^*

1453: \end{equation*}

1454: extends to a $*$-homomorphism from $\Dx$ to $\O_{\OSS}$, where

1455: $S_\mu=S_{\mu_1}S_{\mu_2}\dotsm S_{\mu_k}$ for $\mu=\mu_1\mu_2\dotsm

1456: \mu_k\in\al^*$ with $\mu_1,\mu_2,\dotsc ,\mu_k\in\al$, and

1457: $S_\nu=S_{\nu_1}S_{\nu_2}\dotsm S_{\nu_l}$ for $\nu=\nu_1\nu_2\dotsm

1458: \nu_l\in\al^*$ with $\nu_1,\nu_2,\dotsc ,\nu_l\in\al$. We will

1459: denote this $*$-homomorphism (which in fact is injective) by

1460: $\ioss$.

1461: 

1462: One should notice (cf. \cite{CS}*{Theorem 12}) that when $\OSS$ is a

1463: topological Markov chain with transition matrix $A$

1464: (cf. Section \ref{subsec:CK}), then $\O_{\OSS}$ is equal to the

1465: Cuntz-Krieger algebra $\O_A$ (or to be more precise: to the universal

1466: Cuntz-Krieger algebra considered by an Huef and Raeburn in

1467: \cite{MR1452183}, cf. also \cite{MR561974}). 

1468: 

1469: We will for each $a\in \al$, by $\lambda_a$ denote the map on $\DX$

1470: given by

1471: \begin{equation*}

1472:   \lambda_a(f)(x)=

1473:   \begin{cases}

1474:     f(ax)& \text{if }ax\in \OSS,\\

1475:     0& \text{if }ax\notin \OSS,

1476:   \end{cases}

1477: \end{equation*}

1478: and by $\phi_a$ the map on $\DX$ given by

1479: \begin{equation*}

1480:   \phi_a(f)(x)=

1481:   \begin{cases}

1482:     f(\osh(x))& \text{if }x\in D_a,\\

1483:     0& \text{if }x\notin D_a,

1484:   \end{cases}

1485: \end{equation*}

1486: for $f\in\Dx$ and $x\in \OSS$ (cf. \cite{tmc}*{Proposition 4.3 and

1487:   Lemma 8.2}).

1488: 

1489: Since $(\OSS,\osh)$ is a one-sided symbolic dynamical system, we can

1490: associate to it the discrete partial dynamical system

1491: $(\OSS,\theta_{\OSS},\fg{\al})$ of Definition \ref{onesideddef}. We then

1492: have the following theorem:

1493: 

1494: \begin{theorem} \label{theorem:one} 

1495:   Let $(\OSS,\osh)$ be a one-sided shift space and let

1496:   $(\OSS,\theta_{\OSS},\fg{\al})$ be the discrete partial dynamical system

1497:   associated to $(\OSS,\osh)$ as done in Definition \ref{onesideddef}.

1498:   Then $\csp{\OSS,\theta_{\OSS},\fg{\al}}$ is isomorphic to the

1499:   $\cs$-algebra $\O_{\OSS}$ associated to $(\OSS,\osh)$.

1500: 

1501:   More precisely: if $(D_g)_{g\in\fg\al}$ denotes the domains of

1502:   $\theta_{\OSS}$, $(s_g)_{g\in\fg\al}$ denotes the generators of

1503:   $\csp{\OSS,\theta_{\OSS},\fg{\al}}$, and $\DX$, $\ioss$ and

1504:   $(S_a)_{a\in\al}$ are as above, then we have that

1505:   \begin{equation*}

1506:     \spc\{1_{D_g}\mid g\in\fg\al\}=\DX,

1507:   \end{equation*}

1508:   and that there exists a 

1509:   $*$-isomorphism from $\O_{\OSS}$ to $\csp{\OSS,\theta_{\OSS},\fg{\al}}$

1510:   which maps $\ioss(1_{D_g})$ to $s_gs_g^*$ for every $g\in \fg\al$,

1511:   and $S_a$ to $s_a$ for every $a\in\al$.

1512: \end{theorem}

1513: 

1514: \begin{proof}

1515:   To see that $\spc\{1_{D_g}\mid g\in\fg\al\}=\DX$ notice first that

1516:   it follows from Lemma \ref{lemma:onefund} that

1517:   if $\mu,\nu\in \alwords$, then we have that 

1518:   \begin{equation*}

1519:     \cy\mu\nu=\theta_\nu(D_{\mu\inv})=D_\nu\cap D_{\nu\mu\inv}.

1520:   \end{equation*}

1521:   So $\DX\subseteq \spc\{1_{D_g}\mid g\in\fg\al\}$.

1522:   

1523:   If $A$ is a subset of $\OSS$ such that $1_A\in \DX$, then

1524:   $1_{\theta_a(A)}=\lambda_a(1_A)\in \DX$ and

1525:   $1_{\theta_{a\inv}(A)}=\phi_a(1_A)\in \DX$ for $a\in \al$. So

1526:   since we have that

1527:   \begin{equation*}

1528:     D_g=\theta_{b_1}\circ\theta_{b_2}\circ \dotsb \circ\theta_{b_k}(\OSS)

1529:   \end{equation*}

1530:   for every $g\in \fg{\al}$ written in the reduced form $b_1b_2\dotsm

1531:   b_k$, we have that $\spc\{1_{D_g}\mid g\in\fg\al\}\subseteq \DX$.

1532:   Thus $\DX=\spc\{1_{D_g}\mid g\in\fg\al\}$.

1533:   

1534:   Since $\boomap[(\OSS,\theta_{\OSS},\fg\al)]$ is a Boolean

1535:   homomorphism from $\Bx(\OSS,\theta_{\OSS},\fg\al)$ to the set of

1536:   projections in $\csp{\OSS,\theta_{\OSS},\fg{\al}}$ which maps $D_g$ to

1537:   $s_gs_g^*$, it follows from Lemma \ref{lemma:homo} that there exists

1538:   a $*$-homomorphism $\gamma$ from $\DX$ to

1539:   $\csp{\OSS,\theta_{\OSS},\fg{\al}}$ which maps $1_{D_g}$ to

1540:   $s_gs_g^*$ for every $g\in\fg\al$.

1541:     

1542:   For $g\in\fg\al$ and $a\in\al$, we have that 

1543:   \begin{equation*}

1544:     \begin{split}

1545:       \gamma\left(1_{\theta_a(D_g)}\right)&=\gamma(1_{D_a\cap D_{ag}})\\

1546:       &=s_as_a^*s_{ag}s_{ag}^*\\

1547:       &=s_as_a^*s_{ag}s_{ag}^*s_as_a^*\\

1548:       &=s_as_gs_g^*s_a^*,

1549:     \end{split}

1550:   \end{equation*}

1551:   and that

1552:   \begin{equation*}

1553:     \begin{split}

1554:       \gamma\bigl(1_{\theta_{a\inv}(D_g)}\bigr)

1555:       &=\gamma\bigl(1_{D_{a\inv}\cap D_{a\inv g}}\bigr)\\

1556:       &=s_{a\inv}s_{a\inv}^*s_{a\inv g}s_{a\inv g}^*\\

1557:       &=s_{a\inv}s_{a\inv}^*s_{a\inv g}s_{a\inv g}^*s_{a\inv}s_{a\inv}^*\\

1558:       &=s_{a\inv}s_gs_g^*s_{a\inv}^*.

1559:     \end{split}

1560:   \end{equation*}

1561:   Since $\spc\{1_{D_g}\mid g\in\fg\al\}=\DX$, this shows that

1562:   $\gamma\bigl(\phi_a(f)\bigr)=s_a\gamma(f)s_a^*$ and

1563:   $\gamma\bigl(\lambda_a(f)\bigr)=s_a^*\gamma(f)s_a$ for $f\in\DX$ and

1564:   $a\in\al$. Thus we have that

1565:   \begin{equation*}

1566:     \begin{split}

1567:       \gamma\bigl(1_{\cy\mu\nu}\bigr)

1568:       &=\gamma\bigl(1_{\theta_\nu(D_{\mu\inv})}\bigr)\\

1569:       &= \gamma\bigl(\phi_{\nu_1}\circ\phi_{\nu_2}\dotsb\circ

1570:       \phi_{\mu_k}\circ\lambda_{\mu_l}\circ\dotsb\circ\lambda_{\mu_1}(1)\bigr)\\

1571:       &= s_{\nu_1}s_{\nu_2}\dotsm s_{\nu_l}s_{\mu_l}^*\dotsm

1572:       s_{\mu_1}^*s_{\mu_1}\dotsm s_{\mu_l}s_{\nu_k}^*\dotsm

1573:       s_{\nu_1}^*

1574:     \end{split}    

1575:   \end{equation*}

1576:   for $\mu=\mu_1\mu_2\dotsm \mu_l$ and $\nu=\nu_1\nu_2\dotsm \nu_k$ in

1577:   $\al^*$. Hence it follows from the universal property of $\O_{\OSS}$

1578:   that there is a $*$-homomorphism $\tilde{\gamma}$ from

1579:   $\O_{\OSS}$ to $\csp{\OSS,\theta_{\OSS},\fg{\al}}$ such that

1580:   $\tilde{\gamma}\circ\ioss=\gamma$ (and hence

1581:   $\tilde{\gamma}\bigl(\ioss(1_{D_g})\bigr)=s_gs_g^*$ for every $g\in \fg\al$)

1582:   and $\tilde{\gamma}(S_a)=s_a$ for every $a\in\al$.

1583:   

1584:   Let us now look at $\O_{\OSS}$. Let $S_{a\inv}=S_a^*$ for every

1585:   $a\in\al$, let $S_e=1$ and let for $g\in \fg\al$ written in the

1586:   reduced form $b_1b_2\dotsm b_k$, $S_g=S_{b_1}S_{b_2}\dotsm S_{b_k}$.

1587:   Then $(S_g)_{g\in\fg{\al}}$ is a family of elements from $\O_{\OSS}$ which

1588:   clearly satisfies \eqref{eq:oxa} and \eqref{eq:oxb} of Definition

1589:   \ref{def:ox}. We will show that $(S_g)_{g\in\fg{\al}}$ also

1590:   satisfies \eqref{eq:ox}, \eqref{eq:oxc} and

1591:   \eqref{eq:oxd} of Definition \ref{def:ox}.

1592:   

1593:   It is easy to check that $\ioss\bigl(\lambda_a(f)\bigr)=S_a^*\ioss(f)S_a$

1594:   and $\ioss\bigl(\phi_a(f)\bigr)=S_a\ioss(f)S_{a\inv}$ for $a\in\al$ and

1595:   $f\in \Dx$. Thus if we for $b\in\al\cup\al\inv$ define $\omega_b$ by

1596:   \begin{equation*}

1597:     \omega_b=

1598:     \begin{cases}

1599:       \phi_b&\text{if }b\in\al,\\

1600:       \lambda_{b\inv}&\text{if }b\inv\in\al,

1601:     \end{cases}

1602:   \end{equation*}

1603:   then we have that

1604:   \begin{equation*}

1605:     \begin{split}

1606:       \ioss(1_{D_g})

1607:       &=\ioss\Bigl(1_{\theta_{b_1}\circ\theta_{b_2}\circ\dotsb

1608:         \circ\theta_{b_k}(\OSS)}\Bigr)\\

1609:       &=\ioss\bigl(\omega_{b_1}\circ\omega_{b_2}\circ

1610:       \dotsb\circ\omega_{b_k}(1)\bigr)\\

1611:       &=S_{b_1}S_{b_2}\dotsm S_{b_k}S_{b_k}^*\dotsm S_{b_1}^*\\

1612:       &=S_gS_g^*

1613:     \end{split}

1614:   \end{equation*}

1615:   for every $g\in\fg{\al}$ written in the reduced form $b_1b_2\dotsm b_k$.

1616:   This shows that $(S_g)_{g\in\fg{\al}}$ satisfies \eqref{eq:ox} and

1617:   \eqref{eq:oxd} of Definition \ref{def:ox}.

1618:   

1619:   Let $h,i\in \fg\al$, and let $b_1b_2\dotsm b_k$ be the reduced

1620:   form of $h$ and $b_1'b_2'\dotsm b_{k'}'$ be the reduced form of

1621:   $i$. Consider those $l\in\{1,2,\dotsc,k\}$ for which

1622:   $k+1-l\le k'$ and $b_1'=b_k\inv,

1623:   b_2'=b_{k-1}\inv,\dotsc,b_{k+1-l}'=b_l\inv$, and for which $b_{k+2-l}'\ne

1624:   b_{l-1}'$ if $l\ne 1$ and $k+1-l\ne k'$. Notice that if

1625:   such an $l$ exists, then it is necessarily unique. In this case, we

1626:   let $j=b_lb_{l+1}\dotsm b_k$. If no such $l$ exists,

1627:   then we let $j$ be equal to the neutral element.

1628:   

1629:   Let $h'=hj\inv$ and $i'=ji$. We then have that the reduced form of

1630:   $h'$ is $b_1b_2\dotsm b_{l-1}$, the reduced form of $i'$ is

1631:   $b_{k+2-l}'b_{k+3-l}'\dotsm b_{k'}'$ and the reduced form of

1632:   $hi=h'i'$ is $b_1b_2\dotsm b_{l-1}b_{k+2-l}'b_{k+3-l}'\dotsm

1633:   b_{k'}'$ (if $j=e$, then $h'=h$, $i'=i$ and the reduced form of

1634:   $hi=h'i'$ is $b_1b_2\dotsm b_kb_1'b_2'\dotsm b_{k'}'$). Thus

1635:   $S_{h'}S_{i'}=S_{h'i'}=S_{hi}$, and we have that

1636:   \begin{equation*}

1637:     \begin{split}

1638:       S_hS_i&= S_{h'}S_jS_j^*S_{i'}\\

1639:       &= S_{h'}S_{h'}^*S_{h'}S_jS_j^*S_{i'}\\

1640:       &= S_{h'}S_jS_j^*S_{h'}^*S_{h'}S_{i'}\\

1641:       &= S_hS_h^*S_{h'i'}\\

1642:       &= S_hS_h^*S_{hi}.

1643:     \end{split}

1644:   \end{equation*}

1645:   This shows that $(S_g)_{g\in\fg{\al}}$ satisfies \eqref{eq:oxc}

1646:   of Definition \ref{def:ox}.

1647:   

1648:   Hence it follows from the universal property of

1649:   $\csp{\OSS,\theta_{\OSS},\fg{\al}}$ that there is a $*$-homomorphism

1650:   $\psi$ from $\csp{\OSS,\theta_{\OSS},\fg{\al}}$ to $\O_{\OSS}$ such that

1651:   $\psi(s_g)= S_{b_1}S_{b_2}\dotsm S_{b_k}$ for every $g\in \fg\al$,

1652:   where $g\in\fg\al$ is written in the reduced form $b_1b_2\dotsm b_k$.

1653:   

1654:   We have that $\psi\bigl(\tilde{\gamma}(S_a)\bigr)=\psi(s_a)=S_a$ for every

1655:   $a\in\al$, and since $\O_{\OSS}$ is generated by

1656:   $(S_a)_{a\in\al}$, this shows that

1657:   $\psi\circ\tilde{\gamma}=\Id_{\O_{\OSS}}$.

1658:   

1659:   We also have by Lemma \ref{lemma:onestruc} that

1660:   \begin{equation*}

1661:     \tilde{\gamma}\bigl(\psi(s_g)\bigr)= 

1662:     \tilde{\gamma}\bigl(S_{b_1}S_{b_2}\dotsm S_{b_k}\bigr)= 

1663:     s_{b_1}s_{b_2}\dotsm s_{b_k}=s_g

1664:   \end{equation*}

1665:   for every $g\in \fg\al$,

1666:   where $g\in \fg\al$ is written in the reduced form $b_1b_2\dotsm b_k$,

1667:   and since $\csp{\OSS,\theta_{\OSS},\fg{\al}}$ is generated by

1668:   $(s_g)_{g\in\fg\al}$, this shows that

1669:   $\tilde{\gamma}\circ\psi=\Id_{\csp{\OSS,\theta_{\OSS},\fg{\al}}}$.

1670:     

1671:   Thus $\tilde{\gamma}$ is a $*$-isomorphism from $\O_{\OSS}$ to

1672:   $\csp{\OSS,\theta_{\OSS},\fg{\al}}$ which maps $\ioss(1_{D_g})$ to

1673:   $s_gs_g^*$ for every $g\in \fg\al$ and $S_a$ to $s_a$ for every

1674:   $a\in\al$..

1675: \end{proof}

1676: 

1677: \subsection{Cuntz-Krieger algebras} \label{subsec:CK}

1678: 

1679: Let $\ind$ be an arbitrary index set and let

1680: $A=(A(i,j))_{i,j\in\ind}$ be a matrix with entries in $\{0,1\}$ and

1681: having no zero rows. Exel and Laca have in

1682: \cite{MR2000i:46064} defined a $\cs$-algebra $\O_A$ associated with

1683: $A$. It is the universal $\cs$-algebra generated by a family

1684: $(S_i)_{i\in\ind}$ of partial isometries satisfying the following 4

1685: conditions:     

1686: \begin{align}

1687:   &\label{eq:cka}\\

1688:   &\raisebox{0pt}[38pt][47pt]{\parbox{0.8\textwidth}{for each pair of

1689:       finite subsets $X$ and $Y$ of $\ind$ such that the number

1690:       $A(X,Y,j)$ defined by 

1691:       \begin{equation*}

1692:         A(X,Y,j)=\prod_{x\in X}A(x,j)\prod_{y\in Y}(1-A(y,j))

1693:       \end{equation*}

1694:       is zero for all but a finite number of $j$'s in $\ind$, the

1695:       following equation holds:

1696:       \begin{equation*}

1697:         \prod_{x\in X}S_x^*S_x\prod_{y\in Y}(1-S_y^*S_y) =\sum_{j\in

1698:           \ind}A(X,Y,j)S_jS_j^*,

1699:       \end{equation*}}}\\

1700:   &S_i^*S_i \text{ and } S_j^*S_j \text{ commute for all

1701:   }i,j\in\ind,\label{eq:ckb}\\

1702:   &S_i^*S_j=0, \text{ if }i\ne j\in \ind,\label{eq:ckc}\\

1703:   &S_i^*S_iS_j=A(i,j)S_j, \text{ for all }i,j\in \ind\label{eq:ckd}.

1704: \end{align}

1705: We will call the family $(S_i)_{i\in\ind}$ the generators of $\O_A$.

1706: 

1707: If $A$ is a finite matrix, then the conditions

1708: \eqref{eq:cka}--\eqref{eq:ckd} reduce to the ordinary Cuntz-Krieger

1709: relations, and so $\O_A$ is isomorphic to the usual

1710: Cuntz-Krieger algebra of $A$ \cite{MR561974}, or the universal

1711: Cuntz-Krieger algebra considered in \cite{MR1452183}

1712: (cf. \cite{MR2000i:46064}*{Examples 8.9}). 

1713: 

1714: Let $(\XA,\msh)$ be the topological Markov chain associated to

1715: $A$. That is: $\XA$ is the set defined by 

1716: \begin{equation*}

1717:   \XA=\bigl\{(x_n)_{n\in\No}\in\ind^{\No}\bigm| \forall

1718:   k\in\No:A(x_k,x_{k+1})=1\bigr\},

1719: \end{equation*}

1720: and $\msh:\XA\to\XA$ is the map

1721: \begin{equation*}

1722:   x_0x_1x_2\dotsm\mapsto x_1x_2\dotsm .

1723: \end{equation*}

1724: 

1725: Since $(\XA,\msh)$ is a one-sided symbolic dynamical system, we can

1726: associate to it the discrete partial dynamical system

1727: $(\XA,\theta_\XA,\fg{\ind})$ of Definition \ref{onesideddef}.

1728: We then have the following theorem:

1729: 

1730: \begin{theorem} \label{theorem:CK}

1731:   Let $\ind$ be an arbitrary index set, let

1732:   $A=(A(i,j))_{i,j\in\ind}$ be a matrix with entries in $\{0,1\}$ and

1733:   having no zero rows, and let $(\XA,\theta_\XA,\fg{\ind})$

1734:   be the discrete partial dynamical system defined above. Then

1735:   $\csp{\XA,\theta_\XA,\fg{\ind}}$ is isomorphic to the unitization

1736:   $\widetilde{\O}_A$ of $\O_A$.

1737:  

1738:   More precisely: if $(s_g)_{g\in\fg\ind}$ denotes the generators of

1739:   $\csp{\XA,\theta_\XA,\fg{\ind}}$ and $(S_i)_{i\in\ind}$ denotes the

1740:   generators of $\O_A$, then 

1741:   $\csp{\XA,\theta_\XA,\fg{\ind}}$ is generated by its unit and

1742:   $\{s_i\mid i\in\ind\}$, and there exists a unital

1743:   $*$-isomorphism from $\csp{\XA,\theta_\XA,\fg{\ind}}$ to

1744:   $\widetilde{\O}_A$ which maps $s_i$ to $S_i$ for every $i\in\ind$.

1745: \end{theorem}

1746: 

1747: \begin{proof}

1748:   Let $(D_g)_{g\in\fg\ind}$ and $(\theta_g)_{g\in\fg\ind}$ denote the

1749:   domains and partial one-to-one maps of $\theta_\XA$.

1750: 

1751:   Using the facts that $s_is_i^*s_i=s_i$ for every $i\in\ind$ and

1752:   that the map $D_g\mapsto s_gs_g^*$ extends to a Boolean

1753:   homomorphism from the Boolean algebra $\BA$

1754:   to the Boolean algebra of projections in the unital abelian $\cs$-algebra

1755:   \begin{equation*}

1756:     \cs\Bigl(\bigl\{s_gs_g^*\in\csp{\XA,\theta_\XA,\fg{\ind}}\bigm|

1757:     g\in\fg\ind\bigr\}\Bigr),

1758:   \end{equation*}

1759:   it is easy to check that the family $(s_i)_{i\in\ind}$ of elements

1760:   of $\csp{\XA,\theta_\XA,\fg{\ind}}$ satisfies condition \eqref{eq:cka},

1761:   \eqref{eq:ckb}, \eqref{eq:ckc} and

1762:   \eqref{eq:ckd} above. Thus it follows from the universal property of $\O_A$

1763:   that there is a $*$-homomorphism $\phi$ from $\O_A$ to

1764:   $\csp{\XA,\theta_\XA,\fg{\ind}}$ such that $\phi(S_i)=s_i$ for all

1765:   $i\in\ind$. 

1766: 

1767:   If $\O_A$ is not unital, then $\phi$ extends to a unital $*$-homomorphism

1768:   $\tilde{\phi}$ from $\widetilde{\O}_A$ to

1769:   $\csp{\XA,\theta_\XA,\fg{\ind}}$. If $\O_A$ is unital (in which case

1770:   $\widetilde{\O}_A=\O_A$), then it follows from

1771:   \cite{MR2000i:46064}*{Proposition 8.5} that there exist finite

1772:   subsets $X$ and $Y$ of $\ind$ such that the equation

1773:   \begin{equation*}

1774:     \XA=\bigcup_{x\in X}D_x\cup\bigcup_{y\in Y}D_{y\inv}

1775:   \end{equation*}

1776:   holds.

1777:   That means that the unit of $\csp{\XA,\theta_\XA,\fg{\ind}}$ is

1778:   contained in the $\cs$-subalgebra generated by $\{s_i\mid

1779:   i\in\ind\}$ and thus is in the image of $\phi$, and so $\phi$ is

1780:   unital. 

1781: 

1782:   So we have in both cases that there exists a unital $*$-homomorphism

1783:   $\tilde{\phi}$ from $\widetilde{\O}_A$ to

1784:   $\csp{\XA,\theta_\XA,\fg{\ind}}$ which maps $S_i$ to $s_i$ for all $i\in\ind$.

1785: 

1786:   Let us now turn towards $\widetilde{\O}_A$. We let for every

1787:   $i\in\ind$, $S_{i\inv}=S_i^*$ and we let $S_e=1$ and

1788:   $S_g=S_{b_1}S_{b_2}\dotsm S_{b_k}$, where $g=b_1b_2\dotsm b_k\in

1789:   \fg\ind$ is written in reduced form. It then follows from

1790:   condition \eqref{eq:ckb}, \eqref{eq:ckc} and \eqref{eq:ckd} that the map

1791:   \begin{equation*}

1792:     g\mapsto S_g,\quad g\in \fg\ind

1793:   \end{equation*}

1794:   is a partial representation of $\fg\ind$ (see

1795:   \cite{MR2000i:46064}*{Proposition 3.2} for a proof of this) and

1796:   thus that $(S_g)_{g\in\fg\ind}$ is a family of partial isometries which

1797:   satisfies condition \eqref{eq:ox}--\eqref{eq:oxc} of Definition

1798:   \ref{def:ox}.

1799:   We will now show that $(S_g)_{g\in \fg\ind}$ also satisfies

1800:   \eqref{eq:oxd} of Definition \ref{def:ox}.

1801:   

1802:   It follows from Lemma \ref{lemma:onestruc} that

1803:   $s_g=s_{b_1}s_{b_2}\dotsm s_{b_k}$ for every 

1804:   $g\in \fg\ind$ written in the reduced form $b_1b_2\dotsm b_k$.

1805:   Thus the $*$-homomorphism $\tilde{\phi}:\widetilde{\O}_A\to

1806:   \csp{\XA,\theta_\XA,\fg{\ind}}$ 

1807:   mentioned above maps $S_g$ to $s_g$ for every $g\in \fg\ind$, and

1808:   thus induces a Boolean homomorphism from the Boolean algebra of

1809:   projections in the unital abelian $\cs$-algebra

1810:   \begin{equation*}

1811:     \cs\Bigl(\bigl\{S_g S_g^*\in\widetilde{\O}_A\bigm| g\in\fg\ind\bigr\}\Bigr)

1812:   \end{equation*}

1813:   to the Boolean algebra of projections in the unital abelian $\cs$-algebra

1814:   \begin{equation*}

1815:     \cs\Bigl(\bigl\{s_gs_g^*\in\csp{\XA,\theta_\XA,\fg{\ind}}\bigm|

1816:     g\in\fg\ind\bigr\}\Bigr)

1817:   \end{equation*}

1818:   which maps $S_gS_g^*$ to $s_gs_g^*$. Let us denote the Boolean

1819:   subalgebra of the Boolean algebra of projections in the unital

1820:   abelian $\cs$-algebra 

1821:   \begin{equation*}

1822:     \cs\Bigl(\bigl\{S_g S_g^*\in\widetilde{\O}_A\bigm| g\in\fg\ind\bigr\}\Bigr)

1823:   \end{equation*}

1824:   generated by $\{S_gS_g^*\mid g\in \fg\ind\}$ by $\BO$. Remember

1825:   (cf. Definition 

1826:   \ref{remarkboomap} and Corollary \ref{boomapinj}) that

1827:   $\boomap[(\XA,\theta_\XA,\fg{\ind})]$

1828:   is an injective Boolean homomorphism from $\BA$ to the Boolean algebra

1829:   of projections in the unital abelian $\cs$-algebra

1830:   \begin{equation*}

1831:     \cs\Bigl(\bigl\{s_gs_g^*\in\csp{\XA,\theta_\XA,\fg{\ind}}\bigm|

1832:       g\in\fg\ind\bigr\}\Bigr)

1833:   \end{equation*} 

1834:   which maps $D_g$ to $s_gs_g^*$. Thus there is a Boolean

1835:   homomorphism $\eta$ from $\BO$ to $\BA$ which maps

1836:   $S_gS_g^*$ to $D_g$. We claim that $\eta$ is injective.

1837:   

1838:   To prove this we will use that the family $(S_g)_{g\in\fg\ind}$ has the

1839:   following properties (remember that we by $\ind^*$ denote the set

1840:   of finite words with letters from $\ind$, and that we identify

1841:   $\ind^*$ with the unital sub-semigroup of $\fg\ind$ generated by

1842:   $\ind$, cf. Section \ref{sec:notat}):

1843:   \begin{eqnarray}

1844:     &&S_h^*S_hS_i^*S_i=S_i^*S_iS_h^*S_h \text{ for all

1845:     }h,i\in\fg\ind,\label{eq:ckpa}\\

1846:     &&S_\mu S_\nu=S_{\mu\nu} \text{ for all }\mu,\nu\in\ind^*,\label{eq:ckpb}\\

1847:     &&S_\mu^*S_\nu=0 \text{ if }\mu,\nu\in\ind^*,\

1848:     \abs{\mu}=\abs{\nu}\text{ and }\mu\ne\nu,\label{eq:ckpc}\\

1849:     &&S_\mu^*S_\mu=A(\mu_1,\mu_2)\dotsm

1850:     A(\mu_{\abs{\mu}-1},\mu_{\abs{\mu}})

1851:     S_{\mu_{\abs{\mu}}}^*S_{\mu_{\abs{\mu}}}\text{ for all

1852:     }\mu\in\ind^*,\label{eq:ckpd}\\

1853:     &&\text{if }S_g\ne 0\text{ then }g=\mu\nu\inv \text{ and }S_g=S_\mu

1854:     S_\nu^*\text{ for some

1855:     }\mu,\nu\in\ind^*.\label{eq:ckpe} 

1856:   \end{eqnarray}

1857:   Property \eqref{eq:ckpa} follows from the fact that $g\mapsto S_g$

1858:   is a partial representation of $\fg\ind$, \eqref{eq:ckpb}

1859:   follows from the definition of $S_g$, \eqref{eq:ckpc} is

1860:   \emph{Claim 2} and \eqref{eq:ckpd} is \emph{Claim 1} in the proof

1861:   of \cite{MR2000i:46064}*{Proposition 3.2}. To see \eqref{eq:ckpe}

1862:   notice that it follows from \cite{MR2000i:46064}*{Proposition 3.1}

1863:   that if $S_g\ne 0$, then $g=\mu\nu\inv$ for some $\mu,\nu\in\ind^*$,

1864:   and if we choose $\mu$ and $\nu$ such that the last symbol of $\mu$

1865:   is different from the first letter of $\nu$ (or $\mu$ or $\nu$ is

1866:   the empty word), then $S_g=S_\mu S_\nu^*$ by definition.    

1867:   

1868:   Let for every $\mu\in\ind^*$ and every pair

1869:   $(I,J)$ of (possible empty) finite subsets of $\ind$, $C(\mu,I,J)$

1870:   be the subset of $\XA$ defined by

1871:   \begin{equation*}

1872:     \begin{split}

1873:       C(\mu,I,J)&=\theta_\mu\biggl(\Bigl(\bigcap_{i\in

1874:         I}D_{i\inv}\Bigr)\cap\Bigl(\bigcap_{j\in J}\XA\setminus

1875:       D_{j\inv}\Bigr)\biggr)\\

1876:       &= \bigl\{\mu x\in \XA\bigm| \forall i\in I: ix\in\XA, \forall

1877:       j\in J:jx\notin \XA\bigr\}.

1878:     \end{split}

1879:   \end{equation*}

1880:   Notice that we have that 

1881:   \begin{equation*}

1882:     \eta\Biggl(S_u\prod_{i\in I}S_i^*S_i\prod_{j\in

1883:         J}(1-S_j^*S_j)S_u^*\Biggr)= C(\mu,I,J)

1884:   \end{equation*}

1885:   for finite subsets $I,J$ of $\ind$ and $\mu\in\ind^*$.

1886: 

1887:   We will now prove that every element of $\BA$ can be written as a

1888:   finite union of elements of the form 

1889:   \begin{equation*}

1890:     S_\mu\prod_{i\in I}S_i^*S_i \prod_{j\in J}(1-S_j^*S_j)S_\mu^* \left(

1891:       \prod_{k=1}^n \Biggl(1-S_{\mu^k} \prod_{i\in I_k}S_i^*S_i

1892:         \prod_{j\in J_k}(1-S_j^*S_j)S_{\mu^k}^* \Biggr) \right)

1893:   \end{equation*}

1894:   where $\mu,\mu^1,\mu^2,\dots \mu^n\in \ind^*$ and

1895:   $I,J,I_1,J_1,I_2, J_2, \dotsc ,I_n,J_n$ all are finite (possible

1896:   empty) subsets of $\ind$.

1897: 

1898:   Notice first that if $\mu\in\ind^*$ and $i\in\ind$, then we have that

1899:   \begin{equation*}

1900:     S_\mu S_i^*S_i S_\mu^*=

1901:     \begin{cases}

1902:       S_{i\inv} S_{i\inv}^*&\text{if }u=\emptyword,\\

1903:       S_{\mu i\inv}S_{\mu i\inv}^*&\text{if }\mu_{|\mu|}\ne i,\\

1904:       S_\mu S_\mu^*&\text{if }\mu_{|\mu|}=i,

1905:     \end{cases}

1906:   \end{equation*}

1907:   and thus that $S_\mu S_i^*S_i S_\mu^*$ belongs to $\BO$. Since $\BO$

1908:   is closed under intersection and complement, and the complement of

1909:   an element $p$ of $\BO$ is defined to be $1-p$ and the intersection

1910:   of two elements $p,q\in\BO$ is $pq$, we also have that 

1911:   \begin{equation*}

1912:     S_\mu(1-S_j^*S_j)S_\mu^*=S_\mu S_\mu^*(1-S_\mu S_j^*S_j S_\mu^*)\in\BO

1913:   \end{equation*}

1914:   for every $\mu\in\ind^*$ and every $j\in\ind$.

1915: 

1916:   Let $\mu,\tilde{\mu}\in\ind^*$ and let $I,J,\tilde{I},\tilde{J}$

1917:   be subsets of $I$. It then follows from \eqref{eq:ckd} and

1918:   \eqref{eq:ckpa}--\eqref{eq:ckpc} that the element

1919:   \begin{equation*}

1920:     S_\mu \prod_{i\in I}S_i^*S_i \prod_{j\in J}

1921:     \left(1-S_j^*S_j\right) S_\mu^*

1922:     S_{\tilde{\mu}}\prod_{i\in\tilde{I}}S_i^*S_i \prod_{j\in

1923:       \tilde{J}}\left(1-S_j^*S_j\right) S_{\tilde{\mu}}^*

1924:   \end{equation*}

1925:   is equal to 

1926:   \begin{equation*}

1927:     S_\mu\smashoperator{\prod_{i\in I}}

1928:     S_i^*S_i\smashoperator{\prod_{j\in J}}  

1929:     \left(1-S_j^*S_j\right) S_\mu^*

1930:   \end{equation*} 

1931:   if $\mu=\tilde{\mu}\nu$ for some $\nu=\nu_1\nu_2\dotsm \nu_{|\nu|}

1932:   \in\ind^*$ which satisfies that $A(\tilde{I},\tilde{J},\nu_1)=1$, to

1933:   \begin{equation*}

1934:     S_{\tilde{\mu}}\smashoperator{\prod_{i\in\tilde{I}}}S_i^*S_i

1935:     \smashoperator{\prod_{j\in \tilde{J}}}\left(1-S_j^*S_j\right)

1936:     S_{\tilde{\mu}}^*

1937:   \end{equation*}

1938:   if $\tilde{\mu}=\mu\nu$ for some

1939:   $\nu=\nu_1\nu_2\dotsm \nu_{|\nu|} \in\ind^*$ which satisfies that

1940:   $A(I,J,\nu_1)=1$, to 

1941:   \begin{equation*}

1942:     S_\mu \smashoperator{\prod_{i\in I\cup\tilde{I}}}

1943:     S_i^*S_i \smashoperator{\prod_{j\in J\cup\tilde{J}}} 

1944:     \left(1-S_j^*S_j\right) S_\mu^*

1945:   \end{equation*} 

1946:   if $\mu=\tilde{\mu}$, and to 0 otherwise. 

1947: 

1948:   Thus if we let $\mathcal{Z}$ be the set

1949:   \begin{equation*}

1950:     \Biggl\{S_\mu\prod_{i\in I}S_i^*S_i\prod_{j\in J}

1951:     \left(1-S_j^*S_j\right) S_\mu^*\Bigm| \mu\in\ind^*,\ I,J

1952:     \text{ are finite subset of }\ind^*\Biggr\},

1953:   \end{equation*}

1954:   then $\mathcal{Z}$ is a subset of $\BO$, and it is closed under intersection. 

1955:   So if we let  

1956:   \begin{equation*}

1957:     \mathcal{Y}=\left\{Z_0\bigcap_{Z\in \find}(1-Z)\Bigm| Z_0\in

1958:     \mathcal{Z},\ \find\text{ is a finite subset of }\mathcal{Z} \right\},

1959:   \end{equation*}

1960:   then $\mathcal{Y}$ is also closed under intersection.

1961:   Let $\mathcal{X}$ be the set of

1962:   elements in $\BO$ which can be written as a finite union of 

1963:   elements from $\mathcal{Y}$. Then $\mathcal{X}$ is closed under

1964:   union and intersection, and since we also have that $1-Y$ belongs to

1965:   $\mathcal{X}$ if $Y$ belongs to $\mathcal{Y}$, we also have that

1966:   $\mathcal{X}$ is closed under complement and thus is a Boolean

1967:   subalgebra of $\BO$.

1968:   It follows from \eqref{eq:ckpb}, \eqref{eq:ckpd} and \eqref{eq:ckpe}

1969:   that if $g\in\fg\ind$ and $S_g\ne 0$, then $g=\mu\nu\inv$ for

1970:   suitable $\mu,\nu\in\ind^*$ and that

1971:   \begin{equation*}

1972:     S_g^*S_g=S_\nu S_\mu^*S_\mu S_\nu^*=S_\nu S_i^*S_iS_\nu^*\in\mathcal{X}

1973:   \end{equation*}

1974:   where $i$ is the last letter of $\mu$ (if $\mu$ is the empty word,

1975:   then $S_g^*S_g=S_\nu S_\nu^*\in \mathcal{X}$).

1976:   Thus every element in $\BO$ belongs to $\mathcal{X}$, and is thus a finite

1977:   union of elements of the form

1978:   \begin{equation*}

1979:     S_\mu\prod_{i\in I}S_i^*S_i \prod_{j\in J}(1-S_j^*S_j)S_\mu^* \left(

1980:       \prod_{k=1}^n \Biggl(1-S_{\mu^k} \prod_{i\in I_k}S_i^*S_i

1981:         \prod_{j\in J_k}(1-S_j^*S_j)S_{\mu^k}^* \Biggr) \right)

1982:   \end{equation*}

1983:   where $\mu,\mu^1,\mu^2,\dots \mu^n\in \ind^*$ and

1984:   $I,J,I_1,J_1,I_2, J_2, \dotsc ,I_n,J_n$ all are finite (possible

1985:   empty) subsets of $\ind$.  

1986: 

1987:   In order to prove that $\eta$ is injective, it is therefore enough

1988:   to show that if $\mu,\mu^1,\mu^2,\dots \mu^n\in \ind^*$ and

1989:   $I,J,I_1,J_1,I_2, J_2, \dotsc ,I_n,J_n$ all are finite (possible

1990:   empty) subsets of $\ind$ and 

1991:   \begin{equation*}

1992:     C(\mu,I,J)\cap\left(\bigcap_{k=1}^n \XA\setminus

1993:       C(\mu^k,I_k,J_k)\right)=\emptyset, 

1994:   \end{equation*}

1995:   then we have that

1996:   \begin{equation*}

1997:     S_\mu\prod_{i\in I}S_i^*S_i\prod_{j\in J}(1-S_j^*S_j)S_\mu^*

1998:     \left(\prod_{k=1}^n \Biggl(1-S_{\mu^k}\prod_{i\in I_k}

1999:       S_i^*S_i\prod_{j\in J_k}(1-S_j^*S_j)S_{\mu^k}^*\Biggr)\right)=0.

2000:   \end{equation*}

2001:   

2002:   Let $\mu,\nu\in\ind^*$ with $|\mu|>|\nu|$ (remember that $|\mu|$

2003:   and $|\nu|$ denote the length of $\mu$ and $\nu$, respectively, cf. Section

2004:   \ref{sec:notat}) and let $I,J,I',J'$ be finite subsets of

2005:   $\ind$. Then either the equality 

2006:   \begin{equation*}

2007:     C(\mu,I,J)\cap \XA\setminus C(\nu,I',J')=C(\mu,I,J)

2008:   \end{equation*}

2009:   holds, or $\mu=\nu\alpha$ for some $\alpha=\alpha_1\alpha_2\dotsm

2010:   \alpha_{|\mu|-|\nu|}\in\ind^*$ which satisfies that $A(I',J',\alpha_1)=1$. 

2011:   In the latter case, we have that

2012:   \begin{equation*}

2013:     S_\mu S_\mu^*S_\nu\prod_{i\in I_k}S_i^*S_i

2014:     \prod_{j\in J_k}(1-S_j^*S_j)S_\nu^* 

2015:     = S_\mu S_\mu^*, 

2016:   \end{equation*}

2017:   and thus that

2018:   \begin{equation*}

2019:     S_\mu\prod_{i\in I}S_i^*S_i\prod_{j\in J}(1-S_j^*S_j)S_\mu^*

2020:     \Biggl(1-S_\nu\prod_{i\in I'}S_i^*S_i\prod_{j\in J'}

2021:     (1-S_j^*S_j)S_\nu^*\Biggr)=0.

2022:   \end{equation*}

2023:   So in order to prove that the condition

2024:   \begin{equation*}

2025:     C(\mu,I,J)\cap\left(\bigcap_{k=1}^n \XA\setminus

2026:       C(\mu^k,I_k,J_k)\right)=\emptyset,

2027:   \end{equation*}

2028:   implies that

2029:   \begin{equation*}

2030:     S_\mu\prod_{i\in I}S_i^*S_i\prod_{j\in J}(1-S_j^*S_j)S_\mu^*

2031:     \left(\prod_{k=1}^n \Biggl(1-S_{\mu^k}\prod_{i\in

2032:           I_k}S_i^*S_i\prod_{j\in

2033:           J_k}(1-S_j^*S_j)S_{\mu^k}^*\Biggr)\right)=0,

2034:   \end{equation*}

2035:   we may assume that $|\mu^k|\ge|\mu|$ for every $k\in\{1,2,\dotsc ,n\}$.

2036:  

2037:   So let $\mu,\mu^1,\mu^2,\dots \mu^n\in \ind^*$ with

2038:   $|\mu^k|\ge|\mu|$ for every $k\in\{1,2,\dotsc ,n\}$ and let

2039:   $I,J,I_1,J_1,I_2, J_2, \dotsc ,I_n,J_n$ be finite subsets of $\ind$

2040:   such that

2041:   \begin{equation*}

2042:     C(\mu,I,J)\cap\left(\bigcap_{k=1}^n \XA\setminus

2043:       C(\mu^k,I_k,J_k)\right)=\emptyset.

2044:   \end{equation*}

2045:   

2046:   \begin{clai}

2047:     We claim that there for every $m>|\mu|$ exists a finite subset

2048:     $\fw_m$ of the set

2049:     \begin{multline*}

2050:       \biggl\{\nu\in\ind^*\Bigm| |\nu|=m,\ \forall x\in

2051:       \XA:A(\nu_m,x_0)=1\Longrightarrow \\

2052:       \nu x\in C(\mu,I,J)\cap

2053:       \Bigl(\smashoperator[r]{\bigcap_{\substack{k=1\\

2054:             |\mu^k|<m}}^n} 

2055:       \XA\setminus C(\mu^k,I_k,J_k)\Bigr)\biggr\}

2056:     \end{multline*}

2057:     such that the following equality holds:

2058:     \begin{multline*}

2059:       S_\mu\prod_{i\in I}S_i^*S_i\prod_{j\in J}(1-S_j^*S_j)S_\mu^*

2060:       \Biggl(\prod_{k=1}^n \biggl(1-S_{\mu^k}\prod_{i\in I_k}

2061:       S_i^*S_i\prod_{j\in J_k}(1-S_j^*S_j)S_{\mu^k}^*\biggr)\Biggr)\\

2062:       =\sum_{\nu\in\fw_m}S_\nu S_\nu^* \smashoperator[r]{

2063:         \prod_{\substack{k=1\\

2064:             |\mu^k|\ge m}}^n}

2065:       \biggl(1-S_{\mu^k} \prod_{i\in I_k}S_i^*S_i\prod_{j\in J_k}

2066:       (1-S_j^*S_j)S_{\mu^k}^*\biggr).

2067:     \end{multline*}   

2068:   \end{clai}

2069:   

2070:   It follows from this claim that the equality

2071:   \begin{equation*}

2072:     S_\mu\prod_{i\in I}S_i^*S_i\prod_{j\in J}(1-S_j^*S_j)S_\mu^*

2073:     \left(\prod_{k=1}^n \Biggl(1-S_{\mu^k}\prod_{i\in

2074:         I_k}S_i^*S_i\prod_{j\in

2075:         J_k}(1-S_j^*S_j)S_{\mu^k}^*\Biggr)\right)=0,

2076:   \end{equation*}

2077:   holds, because since $A$ has no identically zero row, there is for every

2078:   $\nu=\nu_1\nu_2\dotsm \nu_m\in\ind^*$, an $x\in\XA$ such that

2079:   $A(\nu_m,x_0)=1$, and it follows from this that the set

2080:   \begin{multline*}

2081:     \biggl\{\nu\in\ind^*\Bigm| |\nu|=m,\ \forall x\in

2082:     \XA:A(\nu_m,x_0)=1\Longrightarrow \\

2083:     \nu x\in C(\mu,I,J)\cap

2084:     \Bigl(\smashoperator[r]{\bigcap_{\substack{k=1\\ |\mu^k|< m}}^n} 

2085:     \XA\setminus C(\mu^k,I_k,J_k)\Bigr)\biggr\}

2086:   \end{multline*}

2087:   is empty for $m>\max\{|\mu^k|\mid k\in \{1,2,\dotsc ,n\}\}$.

2088:   

2089:   To prove the claim, we need a little lemma:

2090: 

2091:   \begin{lemma} \label{modsat}

2092:     Let $R$ be a unital ring, let $k\in\N$ and let 

2093:     $x_1,x_2,\dotsc x_n,\in R$. Then we have that

2094:     \begin{equation*}

2095:       1-\prod_{i=1}^nx_i=\sum_{E\in\cset_n}

2096:       \smashoperator[r]{\prod_{i\in\{1,2,\dotsc n\}\setminus E}}

2097:       x_i\hspace{10pt}\prod_{j\in E}(1-x_j)

2098:     \end{equation*}

2099:     where $\cset_n$ is the set of non-empty subsets of 

2100:     $\{1,2,\dots ,n\}$.

2101:   \end{lemma}

2102:   

2103:   \begin{proof}

2104:     We will prove the lemma by induction. The lemma obviously holds

2105:     for $n=1$. Assume now that the lemma holds for $n=m$. Then we have

2106:     that 

2107:     \begin{equation*}

2108:       \begin{split}

2109:         1-\prod_{i=1}^{m+1}x_i=& \left(1-\prod_{i=1}^mx_i\right)x_{m+1} +

2110:         \prod_{i=1}^mx_i(1-x_{m+1}) +

2111:         \left(1-\prod_{i=1}^mx_i\right)(1-x_{m+1}) \\

2112:         =& \begin{multlined}[t]

2113:           \sum_{E\in\cset_m}\smashoperator[r]{\prod_{i\in\{1,2,\dotsc

2114:               m\}\setminus E}}x_i\hspace{11pt}\prod_{j\in E}(1-x_j)x_{m+1} +

2115:           \prod_{i=1}^mx_i(1-x_{m+1})\\ +

2116:           \sum_{E\in\cset_m}\smashoperator[r]{\prod_{i\in\{1,2,\dotsc

2117:             m\}\setminus E}}x_i\hspace{11pt}\prod_{j\in

2118:           E}(1-x_j)(1-x_{m+1}) 

2119:         \end{multlined}\\

2120:         =& \sum_{E\in\cset_{m+1}}\smashoperator[r]{\prod_{i\in\{1,2,\dotsc

2121:           m+1\}\setminus E}}x_i\hspace{16pt}\prod_{j\in E}(1-x_j).

2122:       \end{split}

2123:     \end{equation*}

2124:   \end{proof}

2125: 

2126:   \begin{proof}[Proof of the claim.]

2127:     We will now prove the claim by induction. First let

2128:     $m=|\mu|+1$. Let $K_\mu$ be the subset of $\{1,2,\dotsc ,n\}$

2129:     defined by

2130:     \begin{equation*}

2131:       K_\mu=\bigl\{k\in \{1,2,\dotsc ,n\}\bigm| \mu^k=\mu\bigr\},

2132:     \end{equation*}

2133:     and let for each $k\in K_\mu$, $\cset_k$ be the set of non-empty subsets of

2134:     $I_k\cup J_k$. We let $\pnes$ be the set of families $(E_k)_{k\in

2135:       K_\mu}$ where for each $k\in K_\mu$, $E_k\in\cset_k$, and we 

2136:     let for $E=(E_k)_{k\in K_\mu}\in \pnes$, $I_E$ be the finite

2137:     subset of $\ind$ defined by

2138:     \begin{equation*}

2139:       I_E=I\cup\{\mu_{|\mu|}\}\cup\bigcup_{k\in K_\mu}(J_k\cap

2140:       E_k)\cup (I_k\setminus E_k),

2141:     \end{equation*}

2142:     and we let $J_E$ be the finite subset of $\ind$ defined by

2143:     \begin{equation*}

2144:       J_E=J\cup\bigcup_{k\in K_\mu}(I_k\cap E_k)\cup 

2145:       (J_k\setminus E_k).

2146:     \end{equation*}

2147:     We then have that if $y\in \XA$ and $A(I_E,J_E,y_0)=1$, then

2148:     \begin{equation*}

2149:       \mu y \in C(\mu,I,J)\cap

2150:       \Bigg(\smashoperator[r]{\bigcap_{\substack{k=1\\

2151:             |\mu^k|=|\mu|}}^n} 

2152:       \XA\setminus C(\mu^k,I_k,J_k)\Bigg)\subseteq

2153:       \smashoperator{\bigcup_{\substack{k=1\\ |\mu^k|>|\mu|}}^n}

2154:       C(\mu^k,I_k,J_k), 

2155:     \end{equation*}

2156:     so if we let

2157:     $\fw_m=\Bigl\{\mu j\Bigm| j\in\ind,\ \exists E\in

2158:     \smashoperator{\prod_{k\in B_\mu}} \cset_k:A(I_E,J_E,j)=

2159:     1\Bigr\}$, then $\fw_m$ is a subset of the set 

2160:     \begin{multline*}

2161:       \biggl\{\nu\in\ind^*\Bigm| |\nu|=m,\ \forall x\in

2162:       \XA:A(\nu_m,x_0)=1\Longrightarrow \\

2163:       \nu x\in C(\mu,I,J)\cap

2164:       \Bigl(\smashoperator[r]{\bigcap_{\substack{k=1\\

2165:             |\mu^k|< m}}^n} \XA\setminus C(\mu^k,I_k,J_k)\Bigr)\biggr\}.

2166:     \end{multline*}

2167:     Since $A$ has no zero row, there is for every $j\in \ind$ an $y\in

2168:     \XA$ such that $y_0=j$. Now if $\mu j\in \fw_m$, then we have, as

2169:     mentioned above, that

2170:     \begin{equation*}

2171:       \mu y\in \smashoperator{\bigcup_{\substack{k=1\\

2172:             |\mu^k|>|\mu|}}^n} C(\mu^k,I_k,J_k),

2173:     \end{equation*}

2174:     which means that 

2175:     \begin{equation*}

2176:       j=y_0\in \bigl\{(\mu^k)_m\bigm| k\in \{1,2,\dotsc,n\},\

2177:       |\mu^k|>|\mu|\bigr\}.

2178:     \end{equation*}

2179:     So $\fw_m$ is finite.

2180: 

2181:     If $|\mu^k|=|\mu|$, but $\mu^k\notin K_\mu$, then

2182:     $S_\mu^*S_{\mu^k}=0$ by \eqref{eq:ckpc}, and so we have that

2183:     \begin{multline*}

2184:       S_\mu\prod_{i\in I}S_i^*S_i\prod_{j\in J}(1-S_j^*S_j)S_\mu^*

2185:       \biggl(1-S_{\mu^k}\prod_{i\in I_k}S_i^*S_i\prod_{j\in J_k}

2186:       (1-S_j^*S_j)S_{\mu^k}^*\biggr) \\

2187:       = S_\mu\prod_{i\in I}S_i^*S_i\prod_{j\in J}(1-S_j^*S_j)S_\mu^*.

2188:     \end{multline*}

2189:     Thus we have that

2190:     \begin{equation*}

2191:       \begin{split}

2192:         \begin{multlined}

2193:           \mathmakebox[9.82cm][l]{S_\mu \prod_{i\in I} S_i^*S_i \prod_{j\in J}

2194:             (1-S_j^*S_j)S_\mu^*}\\ 

2195:           \left( \smashoperator[r]{\prod_{\substack{k=1\\

2196:                   |\mu^k|=|\mu|}}^n} 

2197:             \Biggl( 1-S_{\mu^k} \prod_{i\in I_k} S_i^*S_i \prod_{j\in

2198:               J_k} 

2199:             (1-S_j^*S_j) S_{\mu^k}^* \Biggr) \right)

2200:           =\end{multlined}&\\

2201:         \begin{multlined}        

2202:           \mathmakebox[9.82cm][l]{S_\mu \prod_{i\in I} S_i^*S_i

2203:             \prod_{j\in J} (1-S_j^*S_j) S_\mu^*}

2204:           \\\left( \smashoperator[r]{\prod_{k\in K_\mu}}

2205:             \Biggl(1-S_{\mu^k} \prod_{i\in I_k}

2206:             S_i^*S_i \prod_{j\in J_k} (1-S_j^*S_j) S_{\mu^k}^* \Biggr)

2207:           \right)

2208:           =\end{multlined}&\\

2209:         \begin{multlined}

2210:           \mathmakebox[9.82cm][l]{S_\mu \prod_{i\in I} S_i^*S_i

2211:             \prod_{j\in J} (1-S_j^*S_j) S_\mu^*S_\mu} 

2212:           \\\left( \smashoperator[r]{\prod_{k\in K_\mu}}

2213:             \Biggl(1-\prod_{i\in I_k} S_i^*S_i \prod_{j\in J_k}

2214:             (1-S_j^*S_j) \Biggr) \right) S_\mu^*

2215:           =\end{multlined}&\\

2216:         \begin{multlined}

2217:           \mathmakebox[9.82cm][l]{S_\mu \prod_{i\in I} S_i^*S_i

2218:             \prod_{j\in J} (1-S_j^*S_j) S_\mu^*S_\mu} 

2219:           \\\left( \smashoperator[r]{\prod_{k\in K_\mu}}

2220:             \Biggl( \smashoperator[r]{\sum_{E_k\in \cset_k}} \biggl(

2221:             \smashoperator[r]{\prod_{\substack{i\in (J_k\cap E_k)\\ \cup

2222:                   (I_k\setminus E_k)}}} S_i^*S_i \hspace{-0.7em}

2223:             \smashoperator[r]{\prod_{\substack{j\in (I_k\cap E_k)\\ \cup

2224:                   (J_k\setminus E_k)}}} (1-S_j^*S_j) \biggr) \Biggr) \right)

2225:           S_\mu^*

2226:           =\end{multlined}&\\

2227:         \begin{multlined}

2228:           \mathmakebox[9.82cm][l]{S_\mu \prod_{i\in I} S_i^*S_i

2229:             \prod_{j\in J}(1-S_j^*S_j) S_{\mu_{|\mu|}}^*S_{\mu_{|\mu|}}} \\

2230:           \left( \smashoperator[r]{\prod_{k\in K_\mu}} \Biggl(

2231:             \sum_{E_k\in \cset_k} \biggl(

2232:             \smashoperator[r]{\prod_{\substack{i\in (J_k\cap E_k)\\ \cup

2233:                   (I_k\setminus E_k)}}} S_i^*S_i \hspace{-0.7em}

2234:             \smashoperator[r]{\prod_{\substack{j\in (I_k\cap E_k)\\ \cup

2235:                   (J_k\setminus E_k)}}}(1-S_j^*S_j) \biggr) \Biggr) \right)

2236:         S_\mu^*

2237:         =\end{multlined}&\\

2238:       \begin{multlined}

2239:         \mathmakebox[9.82cm][l]{S_\mu\sum_{E\in \pnes} 

2240:           \left(\prod_{i\in I} S_i^*S_i \prod_{j\in J} (1-S_j^*S_j)

2241:             S_{\mu_{|\mu|}}^*S_{\mu_{|\mu|}}\right)}\\

2242:         \left(

2243:           \prod_{k\in K_\mu}

2244:           \smashoperator[r]{\prod_{

2245:               \substack{i\in (J_k\cap E_k)\\ 

2246:                 \cup(I_k\setminus E_k)}}}

2247:           S_i^*S_i \hspace{-0.7em}

2248:           \smashoperator[r]{\prod_{

2249:               \substack{j\in (I_k\cap E_k)\\ 

2250:                 \cup(J_k\setminus E_k)}}}

2251:           (1-S_j^*S_j) \right) S_\mu^*

2252:         =\end{multlined}&\\

2253:       S_\mu \sum_{E\in \pnes} 

2254:       \left(\smashoperator[r]{\prod_{i\in I_E}} S_i^*S_i

2255:         \smashoperator[r]{\prod_{j\in J_E}}(1-S_j^*S_j) \right) S_\mu^*

2256:       =S_\mu\sum_{E\in \pnes}

2257:       \smashoperator{\sum_{\substack{j\in\ind \\

2258:             A(I_E,J_E,j) = 1}}}S_jS_j^*S_\mu^*=&

2259:       \smashoperator{\sum_{\nu\in \fw_m}}S_\nu S_\nu^*,

2260:       \end{split}

2261:     \end{equation*}

2262:     where the second equality follows from \eqref{eq:ckpa} and the

2263:     facts that $\mu^k=\mu$ for $k\in K_\mu$ and that $S_\mu$ is a

2264:     partial isometry, the third follows from Lemma \ref{modsat}, the

2265:     fourth from the \eqref{eq:ckpd} (if $A(\mu_i,\mu_{i+1})=0$ for some

2266:     $i=\{1,2,\dotsc,|\mu|-1\}$, then $S_\mu=0$ according to

2267:     (\ref{eq:ckd}) and (\ref{eq:ckpb}), and the equality still

2268:     holds), the fifth from the distribute law,

2269:     the sixth from the definition of $I_E$ and $J_E$,

2270:     the seventh from \eqref{eq:ckb} (that $ A(I_E,J_E,j)$ vanishes for

2271:     all but a finite number of $j$'s in $\ind$ follows from the fact

2272:     that $\fw_m$ is finite), and the eight from

2273:     \eqref{eq:ckpb} and the definition of $\fw_m$. Thus we have

2274:     proved the claim in the case where $m=|\mu|+1$.

2275:     

2276:     Assume now that $m>|\mu|$ and that

2277:     \begin{multline*}

2278:       S_\mu\prod_{i\in I}S_i^*S_i\prod_{j\in J}(1-S_j^*S_j)S_\mu^*

2279:       \left(\prod_{k=1}^n \Biggl(1-S_{\mu^k}

2280:         \prod_{i\in I_k}S_i^*S_i

2281:         \prod_{j\in J_k}(1-S_j^*S_j)S_{\mu^k}^*\Biggr)\right)\\

2282:       =\smashoperator{\sum_{\nu\in \fw_m}}S_\nu S_\nu^* 

2283:       \smashoperator{\prod_{\substack{k=1\\ |\mu^k|\ge m}}^n}

2284:       \Biggl(1-S_{\mu^k} \prod_{i\in I_k}

2285:       S_i^*S_i\prod_{j\in J_k}(1-S_j^*S_j)S_{\mu^k}^*\Biggr)

2286:     \end{multline*}

2287:     for some finite subset $\fw_m$ of the set

2288:    \begin{multline*}

2289:      \Biggl\{\nu\in\ind^*\Bigm| |\nu|=m,\ \forall x\in

2290:      \XA:A(\nu_m,x_0)=1\Longrightarrow \\

2291:      \nu x\in C(\mu,I,J)\cap

2292:      \Bigl(\smashoperator[r]{\bigcap_{\substack{k=1\\ |\mu^k|<m}}^n} 

2293:      \XA\setminus C(\mu^k,I_k,J_k)\Bigr)\Biggr\}.

2294:    \end{multline*}

2295:    Let for every $\gamma\in \fw_m$, $K_\gamma$ be the subset of

2296:    $\{1,2,\dotsc ,n\}$ defined by

2297:    \begin{equation*}

2298:      K_\gamma=\bigl\{k\in \{1,2,\dotsc ,n\}\bigm| \mu^k=\gamma\bigr\},

2299:    \end{equation*}

2300:    and let for

2301:    each $k\in K_\gamma$, $\cset_k$ be the set of non-empty subset of

2302:    $I_k\cup J_k$.  We 

2303:    let $\pnes$ be the set of families $(E_k)_{k\in K_\gamma}$ where

2304:    $E_k\in\cset_k$ for

2305:    each $k\in K_\gamma$, and we 

2306:    let for $E=(E_k)_{k\in K_\gamma}\in \pnes$, $I_E$ be the finite

2307:    subset of $\ind$ defined by

2308:    \begin{equation*} 

2309:      I_E=\{\gamma_m\}\cup\bigcup_{k\in K_\gamma}(J_k\cap E_k)\cup

2310:      (I_k\setminus E_k)

2311:    \end{equation*}

2312:    and $J_E$ be the finite subset of $\ind$ defined by

2313:    \begin{equation*}

2314:      J_E=\bigcup_{k\in K_\gamma}(I_k\cap E_k)\cup (J_k\setminus

2315:      E_k).

2316:    \end{equation*}

2317:    We then have that if $y\in \XA$ and $A(I_E,J_E,y_0)=1$,

2318:    then

2319:    \begin{equation*}

2320:      \gamma y \in \smashoperator{\bigcap_{\substack{k=1\\

2321:          |\mu^k|= m}}^n} \XA\setminus C(\mu^k,I_k,J_k),

2322:    \end{equation*}

2323:    and since we have that

2324:     \begin{multline*}

2325:       \gamma\in \fw_m \subseteq\Biggl\{\nu\in\ind^*\Bigm| |\nu|=m,\ \forall

2326:       x\in \XA:A(\nu_m,x_0)=1\Longrightarrow \\

2327:       \nu x\in C(\mu,I,J)\cap

2328:       \Bigl(\smashoperator[r]{\bigcap_{\substack{k=1\\ |\mu^k|< 

2329:           m}}^n} \XA\setminus C(\mu^k,I_k,J_k)\Bigr)\Biggr\},

2330:     \end{multline*}

2331:     it follows that

2332:     \begin{equation*}

2333:       \gamma y\in

2334:       C(\mu,I,J)\cap\Bigg(\smashoperator[r]{\bigcap_{\substack{k=1\\

2335:             |\mu^k|\le m}}^n} 

2336:       \XA\setminus C(\mu^k,I_k,J_k)\Bigg)\subseteq

2337:       \smashoperator{\bigcup_{\substack{k=1\\ |\mu^k|>m}}^n} C(\mu^k,I_k,J_k).

2338:     \end{equation*}

2339:     Let $\fw_\gamma$ be the subset of $\ind$ defined by

2340:     \begin{equation*}

2341:       \fw_\gamma=\biggl\{\gamma j\Bigm| j\in\ind,\ \exists E\in

2342:       \smashoperator{\prod_{k\in K_\gamma}}

2343:       \cset_k:A(X_E,Y_E,j)=1\biggr\}.

2344:     \end{equation*}

2345:     Since $A$ has no zero row, there is for every $j\in \ind$ an $y\in

2346:     \XA$ such that $y_0=j$. Now if $\gamma j\in \fw_\gamma$, then we

2347:     have, as mentioned above, that  

2348:     \begin{equation*}

2349:       \gamma y\in \smashoperator{\bigcup_{\substack{k=1\\

2350:             |\mu^k|>m}}^n} C(\mu^k,I_k,J_k), 

2351:     \end{equation*}

2352:     which means that 

2353:     \begin{equation*}

2354:       j=y_0\in \bigl\{(\mu^k)_m\bigm| k\in \{1,2,\dotsc,n\},\ |\mu^k|>m\bigr\}.

2355:     \end{equation*}

2356:     Thus $\fw_\gamma$ is a finite set.

2357:     Hence if we let $\fw_{m+1}=\bigcup_{\gamma\in \fw_m}\fw_\gamma$, then

2358:     $\fw_{m+1}$ is a finite subset of the set

2359:     \begin{multline*}

2360:       \biggl\{\nu\in\ind^*\Bigm| |\nu|=m+1,\ \forall x\in

2361:       \XA:A(\nu_{m+1},x_0)=1\Longrightarrow \\

2362:       \nu x\in C(\mu,I,J)\cap

2363:       \Bigl(\smashoperator[r]{\bigcap_{\substack{k=1\\

2364:             |\mu^k|< m+1}}^n} 

2365:       \XA\setminus C(\mu^k,I_k,J_k)\Bigr)\biggr\}.

2366:     \end{multline*}

2367: 

2368:     Let $\gamma\in\fw_m$. If $|\mu^k|=m$, but $\mu^k\notin K_\gamma$, then

2369:     $S_\gamma^*S_{\mu^k}=0$ by \eqref{eq:ckpc}, so the equality

2370:     \begin{equation*}

2371:       S_\gamma S_\gamma^*

2372:       \left(1-S_{\mu^k}\prod_{i\in I_k}S_i^*S_i

2373:         \prod_{j\in J_k}(1-S_j^*S_j)S_{\mu^k}^*\right) 

2374:       =   S_\gamma S_\gamma^* 

2375:     \end{equation*}

2376:     holds.

2377: 

2378:     Thus we have that

2379:     \begin{equation*}

2380:       \begin{split}

2381:         S_\gamma S_\gamma^* \Biggl( \smashoperator[r]{\prod_{\substack{k=1\\

2382:               |\mu^k|=m}}^n} \Biggr( 1-S_{\mu^k} \prod_{i\in I_k}

2383:             S_i^*S_i \prod_{j\in J_k} (1-S_j^*S_j) S_{\mu^k}^* \biggr)

2384:         \Biggr)=&\\

2385:         S_\gamma S_\gamma^* \Biggl( \smashoperator[r]{\prod_{k\in

2386:             K_\gamma}} \biggl(1-S_{\mu^k} 

2387:             \prod_{i\in I_k} S_i^*S_i \prod_{j\in J_k} (1-S_j^*S_j)

2388:             S_{\mu^k}^* \biggr) \Biggr)=&\\

2389:         S_\gamma S_\gamma^* S_\gamma \Biggl(

2390:         \smashoperator[r]{\prod_{k\in K_\gamma}} 

2391:         \biggl(1-\prod_{i\in I_k} S_i^*S_i \prod_{j\in J_k}

2392:         (1-S_j^*S_j) \biggr) \Biggr) S_\gamma^*=& \\

2393:         S_\gamma S_\gamma^* S_\gamma \Biggl(

2394:         \smashoperator[r]{\prod_{k\in K_\gamma}} \biggl(

2395:         \smashoperator[r]{\sum_{E_k\in \cset_k}} \Bigl(

2396:         \smashoperator[r]{\prod_{\substack{i\in (J_k\cap E_k)\\ 

2397:               \cup (I_k\setminus E_k)}}}

2398:         S_i^*S_i\hspace{-0.7em} 

2399:         \smashoperator[r]{\prod_{\substack{j\in (I_k\cap E_k)\\ 

2400:               \cup (J_k\setminus E_k)}}}

2401:         (1-S_j^*S_j) \Bigr) \biggr) \Biggr)

2402:         S_\gamma^*=&\\

2403:         S_\gamma S_{\gamma_m}^*S_{\gamma_m} \Biggl( 

2404:         \smashoperator[r]{\prod_{k\in K_\gamma}} 

2405:         \biggl( \sum_{E_k\in \cset_k} \Bigl(

2406:         \smashoperator[r]{\prod_{\substack{i\in (J_k\cap E_k)\\ 

2407:               \cup (I_k\setminus E_k)}}}

2408:         S_i^*S_i \hspace{-0.7em}

2409:         \smashoperator[r]{\prod_{\substack{j\in (I_k\cap E_k)\\ 

2410:               \cup (J_k\setminus E_k)}}}

2411:         (1-S_j^*S_j) \Bigr) \biggr) \Biggr)

2412:         S_\gamma^*=&\\

2413:         S_\gamma

2414:         \Biggl(\smashoperator[r]{\sum_{E\in\pnes}}S_{\gamma_m}^*S_{\gamma_m}

2415:         \smashoperator[l]{\prod_{k\in K_\gamma}} 

2416:         \smashoperator[r]{\prod_{\substack{i\in (J_k\cap E_k)\\ 

2417:               \cup (I_k\setminus E_k)}}}

2418:         S_i^*S_i\hspace{-0.7em}

2419:         \smashoperator[r]{\prod_{\substack{j\in (I_k\cap E_k)\\ 

2420:               \cup (J_k\setminus E_k)}}} 

2421:         (1-S_j^*S_j) \Biggr)

2422:         S_\gamma^*=&\\

2423:         S_\gamma\sum_{E\in\pnes} 

2424:         \Biggl(\prod_{i\in I_E} S_i^*S_i \prod_{j\in J_E} 

2425:         (1-S_j^*S_j) \Biggr) S_\gamma^*=&\\

2426:         S_\gamma \sum_{E\in \pnes}

2427:         \smashoperator{\sum_{\substack{j\in\ind \\

2428:               A(I_E,J_E,j) = 1}}}

2429:         S_jS_j^*S_\gamma^* =&

2430:         \smashoperator{\sum_{\eta\in \fw_\gamma}}S_\eta S_\eta^*,

2431:       \end{split}

2432:     \end{equation*}

2433:     where the second equality follows from \eqref{eq:ckpa}, the

2434:     fact that $\mu^k=\gamma$ 

2435:     for $k\in K_\gamma$ and that $S_\gamma$ is a partial isometry, the third

2436:     follows from Lemma \ref{modsat}, the fourth from \eqref{eq:ckpd}

2437:     (if $A(\gamma_i,\gamma_{i+1})=0$ for some

2438:     $i=\{1,2,\dotsc,|\gamma|-1\}$, then $S_\gamma=0$ according to

2439:     (\ref{eq:ckd}) and (\ref{eq:ckpb}), and the equality

2440:     still holds), the fifth from the distribute law,

2441:     the sixth from the definition of $I_E$ and $J_E$, the seventh

2442:     from \eqref{eq:ckb} (that $ A(I_E,J_E,j)$ vanishes for

2443:     all but a finite number of $j$'s in $\ind$ follows from the fact

2444:     that $\fw_\gamma$ is finite), and the eight from

2445:     \eqref{eq:ckpb} and the definition of $\fw_\gamma$.

2446: 

2447:    Thus we have that

2448:    \begin{equation*}

2449:      \begin{split}

2450:         S_\mu\prod_{i\in I}S_i^*S_i\prod_{j\in J}&(1-S_j^*S_j)S_\mu^*

2451:         \Biggl(\prod_{k=1}^n \biggl(1-S_{\mu^k}\prod_{i\in I_k}

2452:         S_i^*S_i\prod_{j\in J_k}(1-S_j^*S_j)S_{\mu^k}^*\biggr)\Biggr)\\

2453:         =&\smashoperator[r]{\sum_{\gamma\in \fw_m}} 

2454:         S_\gamma S_\gamma^*\hspace{.7ex} \smashoperator{\prod_{\substack{k=1\\ 

2455:               |\mu^k|\ge m}}^n}

2456:         \Biggl(1-S_{\mu^k}\prod_{i\in I_k}S_i^*S_i

2457:         \prod_{j\in J_k}(1-S_j^*S_j)S_{\mu^k}^*\Biggr)\\

2458:         =&\smashoperator[r]{\sum_{\eta\in \fw_{m+1}}} 

2459:         S_\eta S_\eta^*\hspace{.7ex} 

2460:         \smashoperator{\prod_{\substack{k=1\\ |\mu^k|\ge

2461:               m+1}}^n}\hspace{.7ex}

2462:         \Biggl(1-S_{\mu^k} \prod_{i\in I_k}

2463:         S_i^*S_i\prod_{j\in J_k}(1-S_j^*S_j)S_{\mu^k}^*\Biggr),

2464:       \end{split}

2465:     \end{equation*}

2466:     which finalizes the induction proof of the claim.

2467:   \end{proof}

2468:   

2469:   Hence the Boolean homomorphism $\eta$ from $\BO$ to $\BA$ is

2470:   injective, and since it obviously also is surjective, it is

2471:   invertible. Thus there is a Boolean homomorphism from $\BA$ to

2472:   the Boolean algebra of projections in the unital abelian

2473:   $\cs$-algebra 

2474:   \begin{equation*}

2475:     \cs\Bigl(\bigl\{S_gS_g^*\in\widetilde{\O}_A\bigm|

2476:     g\in\fg\ind\bigr\}\Bigr),

2477:   \end{equation*}

2478:   sending $D_g$ to $S_gS_g^*$ for every $g\in\fg\ind$. Hence

2479:   $(S_g)_{g\in\fg{\ind}}$ also satisfies \eqref{eq:oxd} of 

2480:   Definition \ref{def:ox}. So it follows from the universal property of

2481:   $\csp{\XA,\theta_\XA,\fg{\ind}}$ that there is a unital $*$-homomorphism

2482:   $\psi$ from $\csp{\XA,\theta_\XA,\fg{\ind}}$ to $\widetilde{\O}_A$, such that

2483:   $\psi(s_g)= S_{b_1}S_{b_2}\dotsm S_{b_k}$ for $g\in \fg\ind$

2484:   written in the reduced form $b_1b_2\dotsm b_k$.

2485:   

2486:   We have that

2487:   $\psi\bigl(\tilde{\phi}(S_i)\bigr)=\psi(s_i)=S_i$ for every $i\in\ind$ and

2488:   that $\psi\bigl(\tilde{\phi}(1)\bigr)=1$, and since $\widetilde{\O}_A$ is

2489:   generated by $\{S_i\mid i\in\ind\}\cup\{1\}$, this shows that

2490:   $\psi\circ\tilde{\phi}=\Id_{\widetilde{\O}_A}$.

2491:   

2492:   According to Lemma \ref{lemma:onestruc}, $s_g=s_{b_1}s_{b_2}\dotsm

2493:   s_{b_k}$ for every $g\in

2494:     \fg\ind$ written in the reduced form $b_1b_2\dotsm b_k$. Thus we

2495:     have that

2496:   \begin{equation*}

2497:     \tilde{\phi}\bigl(\psi(s_g)\bigr)=\tilde{\phi}(S_{b_1}S_{b_2}\dotsm

2498:     S_{b_k})=s_{b_1}s_{b_2}\dotsm s_{b_k}=s_g

2499:   \end{equation*} for

2500:   every $g\in \fg\ind$ written in the reduced form $b_1b_2\dotsm b_k$,

2501:   and since $\csp{\XA,\theta_\XA,\fg{\ind}}$ is generated by

2502:   $(s_g)_{g\in\fg\ind}$, this shows that

2503:   $\tilde{\phi}\circ\psi=\Id_{\csp{\XA,\theta_\XA,\fg{\ind}}}$.  

2504: 

2505:   Thus $\psi$ is a unital

2506:   $*$-isomorphism from $\csp{\XA,\theta_\XA,\fg{\ind}}$ to

2507:   $\widetilde{\O}_A$ which maps $s_i$ to $S_i$ for every $i\in\ind$,

2508:   and since $\widetilde{\O}_A$ is generated by its unit and

2509:   $\{S_i\mid i\in\ind\}$, $\csp{\XA,\theta_\XA,\fg{\ind}}$ is generated by

2510:   its unit and $\{s_i\mid i\in\ind\}$ .

2511: \end{proof}

2512: 

2513: \section{The ideal structure of $\csp{X,\theta,G}$} \label{sec:ideal}

2514: One of the advantages of having a unified construction of the

2515: $\cs$-algebras associated to one-sided shift spaces, crossed

2516: product of two-sided shift spaces and Cuntz-Krieger algebras is

2517: that it is easy to obtain results which holds for all of these

2518: $\cs$-algebras (and of course other $\cs$-algebras which can be constructed as

2519: $\cs$-algebras of discrete partial dynamical systems) and results which relate

2520: these different kind of $\cs$-algebras to each other.

2521: 

2522: We will in this section for every discrete partial dynamical system

2523: $(X,\theta,G)$ show how a $\theta$-invariant (see the definition of

2524: $\theta$-invariant 

2525: below) subset of $X$ gives raise to an ideal in

2526: $\csp{X,\theta,G}$, and from this construct an injective order

2527: preserving map between certain $\theta$-invariant subsets of $X$ and

2528: ideals of $\csp{X,\theta,G}$. This will enable us to recover some

2529: well-known result about the ideal structure of crossed

2530: product of two-sided shift spaces and Cuntz-Krieger algebras, and will

2531: shade new light on the ideal structure of $\cs$-algebra associated to

2532: one-sided shift spaces.

2533: 

2534: We will also obtain a result which relates the $\cs$-algebras of two

2535: different partial dynamical systems and use this to show that for

2536: two-sided shift spaces having a certain property, the crossed product

2537: of the two-sided shift space is a quotient of the $\cs$-algebra

2538: associated to the corresponding one-sided shift space. This lays the

2539: ground for a description of the $K$-theory of the $\cs$-algebra

2540: associated to the one-sided shift space which is explained in

2541: \cite{tmcseii} and \cite{tmcseiv}. 

2542: 

2543: \begin{definition}

2544:    Let $(X,\theta,G)$ be a discrete partial dynamical system and

2545:    $(\theta_g)_{g\in G}$ the partial one-to-one maps of $\theta$. Then we

2546:    say that a subset $Y$ of $X$ is \emph{$\theta$-invariant} if

2547:    $\theta_g(Y)\subseteq Y$ for all $g\in G$.

2548: \end{definition}

2549:   

2550: Let $(X,\theta,G)$ be a discrete partial dynamical system,

2551: $(D_g)_{g\in G}$ the domains and $(\theta_g)_{g\in G}$ the partial

2552: one-to-one maps of $\theta$, and let $Y$ be an $\theta$-invariant

2553: subset of $X$. Then we also have that $\theta_g(X\setminus Y)\subseteq

2554: X\setminus Y$ for all $g\in G$, so if we for every $g\in G$ by

2555: $\theta_{g|X\setminus Y}$ denote the restriction of $\theta_g$ to

2556: $D_{g\inv}\cap X\setminus Y$, then the triple 

2557: \begin{equation*}

2558:   \left((D_g\cap X\setminus Y)_{g\in G},

2559:     (\theta_{g|X\setminus Y})_{g\in G}\right)

2560: \end{equation*}

2561: is a partial action of $G$ on $X\setminus Y$ which we will denote by 

2562: $\theta_{|X\setminus Y}$. 

2563: Thus $(X\setminus Y,G,\theta_{|X\setminus Y})$ 

2564: is a discrete partial dynamical system.

2565: 

2566: \begin{proposition} \label{shortexact}

2567:   Let $(X,\theta,G)$ be a discrete partial dynamical system and

2568:   let $Y$ be a $\theta$-invariant subset of $X$.

2569:   

2570:   Let $\ideal(Y)$ be the ideal of $\csp{X,\theta,G}$ generated

2571:   by the set

2572:   \begin{equation*}

2573:     \{\boomap(A)\mid A\in \B,\ A\subseteq Y\}.

2574:   \end{equation*}

2575:   Then the quotient $\csp{X,\theta,G}/\ideal(Y)$ is

2576:   isomorphic to $\csp{X\setminus Y,\theta_{|X\setminus Y},G}$.

2577: 

2578:   More precisely: if $\bigl(s_g^X\bigr)_{g\in G}$ denotes the generators of

2579:   $\csp{X,\theta,G}$, and $\bigl(s_g^{X\setminus Y}\bigr)_{g\in G}$ denotes

2580:   the generators of 

2581:   $\csp{X\setminus Y,\theta_{|X\setminus Y},G}$, then the map

2582:   \begin{equation*}

2583:     s_g^X+\ideal(Y)\mapsto s_g^{X\setminus Y}

2584:   \end{equation*}

2585:   extends to a $*$-isomorphism from 

2586:   $\csp{X,\theta,G}/\ideal(Y)$ to $\csp{X\setminus

2587:     Y,\theta_{|X\setminus Y},G}$ which maps

2588:   $\boomap(A)+\ideal(Y)$ to 

2589:   $\boomap[(X\setminus Y,\theta_{|X\setminus Y},G)]

2590:   (A\cap X\setminus Y)$

2591:   for every $A\in \B$.

2592: \end{proposition}

2593: 

2594: \begin{proof}

2595:   Let $(D_g)_{g\in G}$ denote the domains of $\theta$ and let $\BXY$ be

2596:   the Boolean algebra on $X\setminus Y$ generated by

2597:   $\{D_g\cap X\setminus Y\mid g\in G\}$. We claim that the following

2598:   identity holds:

2599:   \begin{equation} \label{ww}

2600:     \BXY=\{A\cap X\setminus Y\mid A\in \B\}.

2601:   \end{equation}

2602:   To see this, notice first that $\{A\cap X\setminus Y\mid A\in \B\}$

2603:   is a Boolean algebra on $X\setminus Y$. Then let $g\in G$. Since

2604:   $D_g\in\B$, we have that 

2605:   \begin{equation*}

2606:     D_g\cap X\setminus Y\in \{A\cap X\setminus Y\mid A\in \B\},

2607:   \end{equation*}

2608:   and therefore that $\BXY\subseteq \{A\cap X\setminus Y\mid A\in \B\}$.

2609: 

2610:   It is easy to check that $\{A\subseteq X\mid A\cap X\setminus Y\in

2611:   \BXY\}$ is a Boolean algebra on $X$, and since we have that 

2612:   \begin{equation*}

2613:     D_g\in\{A\subseteq

2614:     X\mid A\cap X\setminus Y\in \BXY\}

2615:   \end{equation*}

2616:   for every $g\in G$, it follows that $\B\in\{A\subseteq X\mid A\cap

2617:   X\setminus Y\in \BXY\}$,  

2618:   which shows that \eqref{ww} holds.

2619:  

2620:   It follows from (\ref{ww}) that the map 

2621:   \begin{equation*}

2622:     \pi:A\mapsto A\cap X\setminus Y

2623:   \end{equation*}

2624:   is a Boolean homomorphism from $\B$ to $\BXY$ which maps $D_g$ to

2625:   $D_g\cap X\setminus Y$ for every $g\in G$. Thus

2626:   $\boomap[(X\setminus Y,\theta_{|X\setminus Y},G)]\circ\pi$ is

2627:   a Boolean homomorphism from $\B$ to the Boolean algebra of

2628:   projections in the unital abelian $\cs$-algebra

2629:   $\cs\Bigl(\bigl\{s_g^{X\setminus Y}s_g^{X\setminus Y*}\bigm| g\in G\bigr\}\Bigr)$

2630:   which maps $D_g$ to $s_g^{X\setminus Y}s_g^{X\setminus Y*}$ for

2631:   every $g\in G$. So it follows from the universal

2632:   property of $\csp{X,\theta,G}$ that there exists a 

2633:   $*$-homomorphism $\psi$ from $\csp{X,\theta,G}$ to 

2634:   $\csp{X\setminus Y,\theta_{|X\setminus Y},G}$ which maps $s_g^X$ to

2635:   $s_g^{X\setminus Y}$ for every $g\in G$, and $\boomap(A)$ to 

2636:   $\boomap[(X\setminus Y,\theta_{|X\setminus Y},G)]\circ\pi(A)$ 

2637:   for every $A\in\B$. We are done with the proof when we have shown that

2638:   $\ker\psi=\ideal(Y)$. 

2639: 

2640:   If $A\subseteq Y$, then $\pi(A)=\emptyset$ and

2641:   therefore $\psi\bigl(\boomap(A)\bigr)=0$. Thus $\ideal(Y)\subseteq \ker

2642:   \psi$.

2643: 

2644:   The quotient map from $\csp{X,\theta,G}$ to

2645:   $\csp{X,\theta,G}/\ideal(Y)$ induces a Boolean map

2646:   $\widetilde{\pi}$ from the Boolean algebra of projections in the

2647:   unital abelian $\cs$-algebra

2648:   \begin{equation*}

2649:     \cs\Bigl(\bigl\{s_g^Xs_g^{X*}\bigm| g\in G\bigr\}\Bigr)

2650:   \end{equation*}

2651:   to the Boolean algebra of projections in the

2652:   unital abelian $\cs$-algebra

2653:   \begin{equation*}

2654:     \cs\Bigl(\bigl\{s_g^Xs_g^{X*}+\ideal(Y)\in

2655:     \csp{X,\theta,G}/\ideal(Y)\bigm| g\in G\bigr\}\Bigr) 

2656:   \end{equation*}

2657:   which maps $p$ to

2658:   $p+\ideal(Y)$ for every projection $p$ in

2659:   $\cs\Bigl(\bigl\{s_g^Xs_g^{X*}\mid g\in G\bigr\}\Bigr)$. The map

2660:   $\widetilde{\pi}\circ\boomap$ 

2661:   is then a Boolean homomorphism from $\B$ to the Boolean algebra of

2662:   projections in the

2663:   unital abelian $\cs$-algebra

2664:   \begin{equation*}

2665:     \cs\Bigl(\bigl\{s_g^Xs_g^{X*}+\ideal(Y)\in

2666:     \csp{X,\theta,G}/\ideal(Y)\bigm| g\in G\bigr\}\Bigr)

2667:   \end{equation*}

2668:   which maps $D_g$ to $s_g^Xs_g^{X*}+ \ideal(Y)$ for every

2669:   $g\in G$. Since $\widetilde{\pi}\circ\boomap (A)=0$ if $A\subseteq

2670:   Y$, we have that

2671:   $\widetilde{\pi}\circ\boomap$ induces a Boolean homomorphism from

2672:   $\BXY$ to the Boolean algebra of projections in the

2673:   unital abelian $\cs$-algebra

2674:   \begin{equation*}

2675:     \cs\Bigl(\bigl\{s_g^Xs_g^{X*}+\ideal(Y)\in

2676:     \csp{X,\theta,G}/\ideal(Y)\bigm| g\in G\bigr\}\Bigr)

2677:   \end{equation*}

2678:   which maps $D_g\cap

2679:   X\setminus Y$ to $s_g^Xs_g^{X*}+ \ideal(Y)$ for every $g\in

2680:   G$. Thus it follows from the universal property of 

2681:   $\csp{X\setminus Y,\theta_{|X\setminus Y},G}$ that there exists

2682:   a $*$-homomorphism $\tau$ from  

2683:   $\csp{X\setminus Y,\theta_{|X\setminus Y},G}$ to

2684:   $\csp{X,\theta,G}/\ideal(Y)$ which maps $s_g^X$ to

2685:   $s_g^X+\ideal(Y)$ for every $g\in G$.

2686: 

2687:   The $*$-homomorphism $\psi$ from $\csp{X,\theta,G}$ to 

2688:   $\csp{X\setminus Y,\theta_{|X\setminus Y},G}$ induces a 

2689:   $*$-homomorphism $\widetilde{\psi}$ from 

2690:   $\csp{X,\theta,G}/\ker(\psi)$ to 

2691:   $\csp{X\setminus Y,\theta_{|X\setminus Y},G}$ which maps 

2692:   $s_g^X+\ker(\psi)$ to $s_g^{X\setminus Y}$. 

2693:   Thus $\tau\circ\widetilde{\psi}$ is a $*$-homomorphism from \\

2694:   $\csp{X,\theta,G}/\ker(\psi)$ to

2695:   $\csp{X,\theta,G}/\ideal(Y)$ which maps $s_g^X+\ker(\psi)$ to

2696:   $s_g^X+\ideal(Y)$. This shows that $\ker(\psi)\subseteq \ideal(Y)$. 

2697: \end{proof}

2698: 

2699: Let $(X,\theta,G)$ be a discrete partial dynamical system and $Y$ a

2700: $\theta$-invariant subset of $X$. Clearly, the ideal $\ideal(Y)$ from

2701: Proposition \ref{shortexact} only depends of the set 

2702: $\{A\in \B\mid A\subseteq Y\}$.

2703: We will for a subset $Y$ of $X$ call the set

2704: \begin{equation*}

2705:   \bigcup \{A\in \B\mid A\subseteq Y\}

2706: \end{equation*}

2707: for the \emph{$\theta$-admissible core} of $Y$, and we will call $Y$

2708: \emph{$\theta$-admissible} if it is equal to its $\theta$-admissible

2709: core. We then have for every $\theta$-invariant subset $Y$ of $X$ that

2710: the ideal $\ideal(Y)$ from Proposition \ref{shortexact} its identical

2711: to the ideal $\ideal(Y^\circ)$ where $Y^\circ$ denotes the

2712: $\theta$-admissible core of $Y$.

2713: 

2714: Notice that a subset $Y$ of $X$ is $\theta$-admissible if and only if

2715: there for every $x\in Y$ exists a $A\in\B$ such that $x\in A\subseteq Y$.

2716: 

2717: We have just seen how an $\theta$-invariant $\theta$-admissible subset

2718: of $X$ gives raise to an ideal of $\csp{X,\theta,G}$. We will now go

2719: the other way and from an ideal of $\csp{X,\theta,G}$ construct a

2720: $\theta$-invariant $\theta$-admissible subset of $X$.

2721: 

2722: \begin{proposition} \label{prop:invariant}

2723:   Let $(X,\theta,G)$ be a discrete partial dynamical system and let

2724:   $\ideal$ be an ideal in $\csp{X,\theta,G}$. Then the set 

2725:   \begin{equation*}

2726:     \uset=\bigcup\{A\in \B\mid \boomap(A)\in\ideal\}

2727:   \end{equation*}

2728:   is a $\theta$-invariant $\theta$-admissible subset of $X$.

2729: \end{proposition}

2730: 

2731: \begin{proof}

2732:   The set $\uset$ is clearly a $\theta$-admissible subset of $X$, and

2733:   it follows from Lemma \ref{booinvariant} that it is $\theta$-invariant.

2734: \end{proof}

2735: 

2736: \begin{proposition} \label{prop:ideal}

2737:   Let $(X,\theta,G)$ be a discrete partial dynamical system and let

2738:   $Y$ be a $\theta$-invariant $\theta$-admissible subset of $X$. Then

2739:   we have that

2740:   \begin{equation*}

2741:     Y=U\bigl(\ideal(Y)\bigr),

2742:   \end{equation*}

2743:   where $\ideal(Y)$ is as in Proposition \ref{shortexact} and

2744:   $U\bigl(\ideal(Y)\bigr)$ is as in Proposition \ref{prop:invariant}.

2745: \end{proposition}

2746: 

2747: \begin{proof}

2748:   Since $Y$ and $U\bigl(\ideal(Y)\bigr)$ both are $\theta$-admissible

2749:   subsets of $X$, it is enough to prove that the equivalence

2750:   \begin{equation*}

2751:     A\subseteq Y\iff \boomap(A)\in \ideal(Y)

2752:   \end{equation*}

2753:   holds for all $A\in \B$. It is clear that $\boomap (A)\in \ideal(Y)$ if

2754:   $A\subseteq Y$. 

2755: 

2756:   Assume that $A$ is not a subset of $Y$. Then $A\cap X\setminus Y$ is

2757:   non-empty, and since 

2758:   $\boomap[(X\setminus U,\theta_{|X\setminus Y},G)]$ is injective, 

2759:   $\boomap[(X\setminus Y,\theta_{|X\setminus Y},G)]

2760:   (A\cap X\setminus Y)$ is a non-zero element of 

2761:   $\csp{X\setminus Y,\theta_{|X\setminus Y},G}$. 

2762:   By Proposition \ref{shortexact}, this

2763:   implies that $\boomap(A)+\ideal(Y)$ is non-zero,

2764:   and thus that $\boomap(A)$ is not in $\ideal(Y)$. 

2765: \end{proof}

2766: 

2767: From Proposition \ref{prop:ideal} now directly follows the promised

2768: theorem about the existence of  an injective order

2769: preserving map between certain $\theta$-invariant subsets of $X$ and

2770: ideals of $\csp{X,\theta,G}$: 

2771: 

2772: \begin{theorem} \label{theorem:ideal}

2773:   Let $(X,\theta,G)$ be a discrete partial dynamical system and let

2774:   for every  $\theta$-invariant $\theta$-admissible subset $Y$ of

2775:   $X$, $\ideal(Y)$ be the ideal of $\csp{X,\theta,G}$ generated

2776:   by the set 

2777:   \begin{equation*}

2778:     \{\boomap(A)\mid A\in \B,\ A\subseteq Y\}.

2779:   \end{equation*}

2780:   Then the map

2781:   \begin{equation*}

2782:     Y\mapsto \ideal(Y)

2783:   \end{equation*}

2784:   is an injective order preserving (i.e., $Y\subseteq Z\Rightarrow

2785:   \ideal(Y)\subseteq \ideal(Z)$) map from the set of $\theta$-invariant

2786:   $\theta$-admissible subsets of $X$ to the set of ideals of

2787: \end{theorem}

2788: 

2789: We will now present a result which relate the $\cs$-algebras of two

2790: partial dynamical systems to each other:

2791: 

2792: \begin{theorem} \label{theorem:surmap}

2793:   Let $G$ be a group and let for $i\in\{1,2\}$, $X_i$ be a set,

2794:   $\theta_i$ a partial action of $G$ on $X_i$ and

2795:   $(\theta_g^{X_i})_{g\in G}$ the partial one-to-one maps of

2796:   $\theta_i$. If there exists a 

2797:   Boolean homomorphism $\boof$ from $\Bx(X_1,\theta_1,G)$ to

2798:   $\Bx(X_2,\theta_2,G)$ such that the identity

2799:   \begin{equation*}

2800:     \boof\bigl(\theta_g^{X_1}(A)\bigr)=\theta_g^{X_2}\bigl(\boof(A)\bigr)

2801:   \end{equation*}

2802:   holds for all $A\in \Bx(X_1,\theta_1,G)$ and $g\in G$, then

2803:   $\csp{X_2,\theta_2,G}$ is a quotient of $\csp{X_1,\theta_1,G}$.

2804:  

2805:   More precisely: if for $i\in\{0,1\}$, $(s_g^{X_i})_{g\in G}$ denotes the

2806:   generators of $\csp{X_i,\theta_i,G}$,

2807:   then there exists a surjective $*$-homomorphism from

2808:   $\csp{X_1,\theta_1,G}$ to $\csp{X_2,\theta_2,G}$ which maps $s_g^{X_1}$ to

2809:   $s_g^{X_2}$ for every $g\in G$, and $\boomap[(X_1,\theta_1,G)](A)$ to

2810:   $\boomap[(X_2,\theta_2,G)]\bigl(\boof(A)\bigr)$ for every

2811:   $A\in\Bx(X_1,\theta_1,G)$. 

2812: 

2813:   The kernel of this $*$-homomorphism is

2814:   the ideal generated by the set

2815:   \begin{equation*}

2816:     \{\boomap[(X_1,\theta_1,G)](A)\mid A\in\Bx(X_1,\theta_1,G),\

2817:     \boof(A)=\emptyset\}.

2818:   \end{equation*}

2819: \end{theorem}

2820: 

2821: \begin{proof}

2822:   Notice that $Y=\bigcup\{A\in\Bx(X_1,\theta_1,G)\mid

2823:   \boof(A)=\emptyset\}$ is a $\theta_1$-invariant

2824:   $\theta_1$-admissible subset of $X_1$ and that 

2825:   if we let $\ideal(Y)$ be as in Proposition \ref{shortexact}, then

2826:   $\ideal(Y)$ is the ideal generated by the set

2827:   \begin{equation*}

2828:     \{\boomap[(X_1,\theta_1,G)](A)\mid A\in\Bx(X_1,\theta_1,G),\

2829:     \boof(A)=\emptyset\}. 

2830:   \end{equation*}

2831: 

2832:   It directly follows from the universal property of

2833:   $\csp{X_1,\theta_1,G}$ that there exists a $*$-homomorphism $\psi$ from

2834:   $\csp{X_1,\theta_1,G}$ to $\csp{X_2,\theta_2,G}$ which maps $s^{X_1}_g$ to

2835:   $s^{X_2}_g$ for every $g\in G$ and $\boomap[(X_1,\theta_1,G)](A)$ to

2836:   $\boomap[(X_2,\theta_2,G)](\boof(A))$ for every

2837:   $A\in\Bx(X_1,\theta_1,G)$, and it is clear that the set

2838:   \begin{equation*}

2839:     \{\boomap[(X_1,\theta_1,G)](A)\mid A\in\Bx(X_1,\theta_1,G),\

2840:     \boof(A)=\emptyset\} 

2841:   \end{equation*}

2842:   is contained in the kernel of $\psi$. Since $\psi$ maps $s^{X_1}_g$ to

2843:   $s^{X_2}_g$, $\psi$ is surjective, so we only have to show that

2844:   the kernel of $\psi$ is contained in $\ideal(Y)$. We will do that by

2845:   proving that there exists a $*$-homomorphism 

2846:   from $\csp{X_1,\theta_1,G}/\ker\psi$ to $\csp{X_1,\theta_1,G}/\ideal(Y)$

2847:   which maps $s^{X_1}_g+\ker\psi$ to $s^{X_1}_g+\ideal(Y)$ for every $g\in G$.

2848:   

2849:   Let us denote the generators of

2850:   $\csp{X_1\setminus Y,\theta_{1|X_1\setminus Y},G}$ by

2851:   $(s^{X_1\setminus Y}_g)_{g\in G}$.  

2852:   Since $\psi$ is surjective, it induces a $*$-isomorphism from $

2853:   \csp{X_1,\theta_1,G}/\ker\psi$ to $ \csp{X_2,\theta_2,G}$ which maps

2854:   $s^{X_1}_g+\ker\psi$ to $s^{X_2}_g$ and it follows from Proposition

2855:   \ref{shortexact} 

2856:   that there exists a $*$-isomorphism from $

2857:   \csp{X_1,\theta_1,G}/\ideal(Y)$ to $\csp{X_1\setminus

2858:     Y,\theta_{1|X_1\setminus Y},G}$ which maps $s^{X_1}_g+\ideal(Y)$

2859:   to $s^{X_1\setminus Y}_g$

2860:   for every $g\in G$. So all we have to do is to show that there

2861:   exists a $*$-homomorphism from $\csp{X_2,\theta_2,G}$ to

2862:   $\csp{X_1\setminus Y,\theta_{1|X_1\setminus Y},G}$ which maps

2863:   $s^{X_2}_g$ to $s^{X_1\setminus Y}_g$ for every $g\in G$.

2864: 

2865:   Let $(D_g^{X_1})_{g\in G}$ denote the domains of $\theta_1$ and

2866:   $(D_g^{X_2})_{g\in G}$ denote the domains of $\theta_2$. We then have

2867:   that

2868:   \begin{equation*}

2869:     \boof\bigl(D_g^{X_1}\bigr)=\boof\bigl(\theta_g^{X_1}(X_1)\bigr)=

2870:     \theta_g^{X_2}\bigl(\boof(X_1)\bigr) 

2871:     =\theta_g^{X_2}\bigl(X_2\bigr)= D_g^{X_2}

2872:   \end{equation*}

2873:   for every $g\in G$, $\boof$ is

2874:   surjective, and if  $A,B\in\Bx(X_1,\theta_1,G)$ and

2875:   $\boof(A)=\boof(B)$, then $A\cap X_1\setminus Y=B\cap X_1\setminus

2876:   Y$. So the map  

2877:   \begin{equation*}

2878:     \boof(A)\mapsto A\cap X_1\setminus Y

2879:   \end{equation*}

2880:   is a well-defined map from $\Bx(X_2,\theta_2,G)$ to

2881:   $\Bx(X_1\setminus Y,\theta_{1|X_1\setminus Y},G)$ which maps $D_g$

2882:   to $D_g\cap X_1\setminus Y$ for every $g\in G$. It is easy to check

2883:   that it is a Boolean homomorphism, so it follows from the universal

2884:   property of $\csp{X_2,\theta_2,G}$ that there  exists a

2885:   $*$-homomorphism from $\csp{X_2,\theta_2,G}$ to 

2886:   $\csp{X_1\setminus Y,\theta_{1|X_1\setminus Y},G}$ which maps

2887:   $s^{X_2}_g$ to $s^{X_1\setminus Y}_g$ for every $g\in G$.

2888: \end{proof}

2889: 

2890: 

2891: \subsection{The ideal structure of $C(\TSS)\rtimes_{\tsh^\star}\Z$} 

2892: Let $(\TSS,\tsh)$ be a two-sided shift space over the finite

2893: alphabet $\al$. As it is proved in the proof of

2894: Theorem \ref{kryds}, the Boolean algebra

2895: $\Bx(\TSS,\theta_\TSS,\fg\al)$ is equal the Boolean algebra of clopen subsets

2896: of $\TSS$, and since the clopen subsets generate the topology of

2897: $\TSS$, a subset of $\TSS$ is $\theta_\TSS$-admissible if and only it is

2898: open. It is easy to check that a subset $Y$ of $\TSS$ is

2899: $\theta_\TSS$-invariant if and only if $\tsh(Y)=Y$. 

2900: 

2901: Thus we recover from Theorem \ref{kryds} and \ref{theorem:ideal} the

2902: well know fact 

2903: that there exists an injective order preserving map from the set of

2904: open $\tsh$-invariant subsets of $\TSS$ to the set of ideals of

2905: $C(\TSS)\rtimes_{\tsh^\star}\Z$. If $(\TSS,\tsh)$ is free (meaning that

2906: $\tsh^n(x)\ne x$ for every $x\in\TSS$ and every $n\in\Z\setminus\{0\}$)

2907: then this map

2908:  is bijective, cf. \cite{MR0241994}*{Proposition 5.10 and

2909:   Th\'eor\`eme 5.15}.

2910: 

2911: 

2912: \subsection{The ideal structure of $\widetilde{\O}_A$}

2913: \label{subsec:idealkryds} 

2914: Let $\ind$ be an arbitrary index set and let

2915: $A=(A(i,j))_{i,j\in\ind}$ be a matrix with entries in $\{0,1\}$ and

2916: having no zero rows. There is, as mentioned in Remark \ref{remark:exellaca},

2917: a homeomorphism from $\h{\mathsf{X}}_A^+$ to

2918: $\widetilde{\Omega}_A$. Thus there is a bijective correspondence

2919: between elements of $\BA$ and clopen subsets of

2920: $\widetilde{\Omega}_A$. This correspondence extend to a correspondence

2921: between $\theta_\XA$-admissible subsets of $\XA$ and open subsets of

2922: $\widetilde{\Omega}_A$, and this correspondence takes

2923: $\theta_\XA$-invariant subsets to invariant subsets of

2924: $\widetilde{\Omega}_A$. 

2925: 

2926: Thus we get from Theorem \ref{theorem:CK} and \ref{theorem:ideal}

2927: an injective order preserving map from the set of open invariant

2928: subsets of $\widetilde{\Omega}_A$ to the set of ideals of

2929: $\widetilde{\O}_A$. Exel and Laca have in

2930: \cite{MR2000i:46064}*{Theorem 15.1} proved that this map is bijective

2931: if the directed graph $Gr(A)$ of $A$ has no transitory circuits

2932: (cf. \cite{MR2000i:46064}*{Section 12}).   

2933: 

2934: 

2935: \subsection{The ideal structure of $\O_{\OSS}$} 

2936: Let $(\OSS,\osh)$ be a one-sided shift space over the finite alphabet

2937: $\al$, and let $\laOSS$ be the language of $\OSS$

2938: (cf. \cite{MR97a:58050}*{\S1.3}), that is 

2939: \begin{equation*}

2940:   \laOSS=\{\mu\in\al^*\mid \exists x\in\OSS,0\le k\le

2941:   l:x_kx_{k+1}\dotsm x_l=\mu\}.

2942: \end{equation*}

2943: We let $(\OSS,\theta_{\OSS},\fg{\al})$ be the discrete partial

2944: dynamical system of Theorem \ref{theorem:one}, and 

2945: $(D_g)_{g\in\fg\al}$ be the domains and $(\theta_g)_{g\in\fg\al}$ the

2946: partial one-to-one maps of $\theta_{\OSS}$.

2947: 

2948: Following \cite{MR1691469}, we let for every $x\in \OSS$ and every

2949: $k\in \No$, $\Past_k(x)$ be the subset of $\laOSS$ defined by

2950: \begin{equation*}

2951:   \Past_k(x)=\{\mu\in \laOSS\mid \mu x\in \OSS, |\mu|=k\},

2952: \end{equation*}

2953: and define for every $l\in \No$ an equivalence relation $\sim_l$ on

2954: $\OSS$ by

2955: \begin{equation*}

2956:   x\sim_l x' \iff \Past_l(x)=\Past_l(x').

2957: \end{equation*}

2958: 

2959: We then define for every $(k,l)\in \indd$ an equivalence relation

2960: $\eq$ on $\OSS$ by

2961: \begin{equation*}

2962:   x\eq y\iff x_{[0,k[}=y_{[0,k[}\land

2963:   \Past_l(x_{[k,\infty[})=\Past_l(y_{[k,\infty[}), 

2964: \end{equation*}

2965: and we let $\ec{k}{x}{l}$ denote the equivalence class of $x$ under

2966: this equivalence relation. We then define an order $\lee$ on $\indd$ by

2967: \begin{equation*}

2968:   (k_1,l_1)\lee (k_2,l_2) \iff k_1\le k_2 \land l_1-k_1\le l_2-k_2,

2969: \end{equation*}

2970: and notice that if $ (k_1,l_1)\lee (k_2,l_2)$, then

2971: $\ec{k_2}{x}{l_2}\subseteq\ec{k_1}{x}{l_1}$ for all $x\in\OSS$. Notice also that

2972: $\ec{k}{x}{l}\in\Bx(\OSS,\theta_{\OSS},\fg\al)$ for all $x\in\OSS$ and

2973: all $(k,l)\in\indd$ because we have that

2974: \begin{equation*}

2975:   \ec{k}{x}{l}=\theta_{x_{[0,k[}}

2976:   \left(\Biggl(\smashoperator[r]{\bigcap_{\mu\in\Past_l(x_{[k,\infty[})}}

2977:     (D_{\mu\inv})\Biggr) \cap \Biggl(

2978:   \smashoperator[r]{\bigcap_{\mu\in\neg\Past_l(x_{[k,\infty[})}}

2979:   \OSS\setminus D_{\mu\inv}\Biggr)\right),

2980: \end{equation*}

2981: where $\neg\Past_l(x_{[k,\infty[})=\{\mu\in\laOSS\mid |\mu|=l,\

2982: \mu\notin\Past_l(x_{[k,\infty[})\}$.

2983: 

2984: \begin{lemma} \label{lemma:clopen}

2985:   There exists for every $A\in\Bx(\OSS,\theta_{\OSS},\fg\al)$ a

2986:   $(k,l)\in\indd$ such that the implication

2987:   \begin{equation*}

2988:     x\in A\Rightarrow \ec{k}{x}{l}\subseteq A

2989:   \end{equation*}

2990:   holds for all $x\in\OSS$.

2991: \end{lemma}

2992: 

2993: \begin{proof}

2994:   Let $\Aa$ be the subset of $\Bx$ defined by

2995:   \begin{equation*}

2996:     \Aa=\bigl\{A\in\Bx(\OSS,\theta_{\OSS},\fg\al)\bigm| \exists

2997:     (k,l)\in\indd\forall x\in A:\ec{k}{x}{l}\subseteq A\bigr\}. 

2998:   \end{equation*}

2999:   We must then show that $\Aa=\Bx(\OSS,\theta_{\OSS},\fg\al)$. Clearly

3000:   $\OSS\in \Aa$. 

3001:   Assume that $A,B\in \Aa$ and choose $(k_a,l_a),(k_b,l_b)\in \indd$

3002:   such that the two implications

3003:   \begin{equation*}

3004:     x\in A\Rightarrow \ec{k_a}{x}{l_a}\subseteq A

3005:   \end{equation*}

3006:   and

3007:   \begin{equation*}

3008:     x\in B\Rightarrow \ec{k_b}{x}{l_b}\subseteq B

3009:   \end{equation*}

3010:   hold for all $x\in\OSS$.

3011:   

3012:   Let $k=\max\{k_a,k_b\}$ and $l=\max\{l_a-k_a,l_b-k_b\}+k$. Then we

3013:   have that $(k_a,l_a),(k_b,l_b)\lee (k,l)$, from which it follows

3014:   that the implication 

3015:   \begin{equation*}

3016:     x\in A\cap B\Rightarrow

3017:     \ec{k}{x}{l}\subseteq\ec{k_a}{x}{l_a}\cap\

3018:     \ec{k_b}{x}{l_b}\subseteq A\cap B,

3019:   \end{equation*}

3020:   holds for all $x\in\OSS$. This shows that $A\cap B\in \Aa$. We

3021:   also have that the three implications

3022:   \begin{gather*}

3023:     x\in \OSS\setminus A\Rightarrow \ec{k_a}{x}{l_a}\subseteq 

3024:     \OSS\setminus A,\\

3025:     x\in \theta_a(A)\Rightarrow \ec{k_a+1}{x}{l_a}\subseteq\theta_a(A)

3026:   \end{gather*}

3027:   and

3028:   \begin{equation*}

3029:     x\in \theta_{a\inv}(A)\Rightarrow \ec{k_a}{x}{l_a+1}\subseteq

3030:     \theta_{a\inv}(A) 

3031:   \end{equation*}

3032:   hold for all $x\in\OSS$. Thus the sets $\OSS\setminus A,\

3033:   \theta_a(A)$, and $\theta_{a\inv}(A)$ all belong to $\Aa$. Hence

3034:   $\Aa$ is a Boolean algebra containing $D_g$ for every 

3035:   $g\in \fg\al$, which means that $\Aa=\Bx(\OSS,\theta_{\OSS},\fg\al)$.

3036: \end{proof}

3037: 

3038: \begin{lemma} \label{lemma:open}

3039:   A subset $Y$ of $\OSS$ is $\theta_{\OSS}$-admissible if and only if there

3040:   for every $x\in Y$ exists a $(k,l)\in\indd$ such that

3041:   $\ec{k}{x}{l}\subseteq Y$.  

3042: \end{lemma}

3043: 

3044: \begin{proof}

3045:   Let $\Aa$ be the subset of $\OSS$ defined by 

3046:   \begin{equation*}

3047:     \Aa=\bigl\{Y\subseteq \OSS\bigm| \forall x\in Y\exists

3048:     (k,l)\in\indd: \ec{k}{x}{l}\subseteq Y\bigr\}.

3049:   \end{equation*}

3050:   We will show that a subset $Y$ of $\OSS$ belongs to $\Aa$ if and

3051:   only if it is $\theta_{\OSS}$-admissible.

3052: 

3053:   It is clear that if $(Y_i)_{i\in\ind}$ is a family of elements of

3054:   $\Aa$, then $\bigcup_{i\in\ind}Y_i\in \Aa$, and since it follows

3055:   from Lemma \ref{lemma:clopen} that 

3056:   $\Bx(\OSS,\theta_{\OSS},\fg\al)\subseteq\Aa$, we have that every

3057:   $\theta_{\OSS}$-admissible subset of $\OSS$ belongs to $\Aa$.

3058: 

3059:   Now let $Y\in\Aa$. It is in order to show that $Y$ is

3060:   $\theta_{\OSS}$-admissible enough to show that there for every

3061:   $x\in Y$ exists a $A\in\Bx(\OSS,\theta_{\OSS},\fg\al)$ such that

3062:   $x\in A\subseteq Y$, but we have that $x\in\ec{k}{x}{l}\subseteq Y$,

3063:   and $\ec{k}{x}{l}\in\Bx(\OSS,\theta_{\OSS},\fg\al)$ for some 

3064:   $(k,l)\in\indd$, so we are done. 

3065: \end{proof}

3066: 

3067: It is not difficult to see that a subset $Y$ of $\OSS$ is

3068: $\theta_{\OSS}$-invariant if and only if $\osh(Y)\subseteq Y$ and

3069: $\osh\inv(Y)\subseteq Y$. Combining this with Theorem

3070: \ref{theorem:one}, Theorem \ref{theorem:ideal} and Lemma

3071: \ref{lemma:open} we get: 

3072: 

3073: \begin{theorem} \label{theorem:ossideal}

3074:   Let $(\OSS,\osh)$ be a one-sided shift space over the finite alphabet

3075:   $\al$, let $\mathcal{H}(\OSS)$ denote the set of subsets $Y$ of

3076:   $\OSS$ which satisfies the following 3 conditions: 

3077: \begin{enumerate}

3078:   \item $\osh(Y)\subseteq Y$,

3079:   \item $\osh\inv(Y)\subseteq Y$,

3080:   \item $\forall x\in Y\exists (k,l)\in\indd: \ec{k}{x}{l}\subseteq

3081:     Y$,

3082: \end{enumerate}

3083: and let for every subset $Y$ in $\mathcal{H}(\OSS)$, $\ideal(Y)$ be the

3084: ideal of $\csp{\OSS,\theta_{\OSS},\fg\al}$ generated

3085: by the set

3086: \begin{equation*}

3087:   \{\boomap[(\OSS,\theta_{\OSS},\fg\al)](A)\mid A\in

3088:   \Bx(\OSS,\theta_{\OSS},\fg\al),\ A\subseteq Y\}.

3089: \end{equation*}  

3090: Then the map

3091: \begin{equation*}

3092:   Y\mapsto \ideal(Y)

3093: \end{equation*}

3094: is an injective order preserving (i.e., $ Y\subseteq Z\Rightarrow

3095: \ideal(Y)\subseteq \ideal(Z)$) map from $\mathcal{H}(\OSS)$ to the

3096: set of ideals of $\O_{\OSS}$.

3097: \end{theorem}

3098: 

3099: If $Y$ is an open subset of $\OSS$ then there exists for every $x\in Y$

3100: a $(k,l)\in\indd$ such that $\ec{k}{x}{l}\subseteq Y$ (in fact, one

3101: can choose $l$ to be $0$). There may on the

3102: other hand be subsets $Y$ of $\OSS$ which are not open, but with the

3103: property that $\forall x\in Y\exists (k,l)\in\indd:

3104: \ec{k}{x}{l}\subseteq Y$. If however $\OSS$ is a shift of finite

3105: type (cf. \cite{MR97a:58050}*{\S2.1}), then this can not happen (in

3106: fact according to \cite{MR0346134}*{Theorem 1}, $\OSS$ is of finite

3107: type if and only if $\osh$ is an open map, and it is not difficult to

3108: show that this is equivalent to $\ec{k}{x}{l}$ being clopen for all

3109: $x\in\OSS$ and all $(k,l)\in\indd$). Thus if $\OSS$ is a shift of finite

3110: type, then the set $\mathcal{H}(\OSS)$ from Theorem

3111: \ref{theorem:ossideal} is the set of all open subsets $Y$ of

3112: $\OSS$ with the property that $\osh(Y)\subseteq Y$ and 

3113: $\osh\inv(Y)\subseteq Y$. Hence we

3114: get the following corollary:

3115: 

3116: \begin{corollary}

3117:   Let $(\OSS,\osh)$ be a one-sided shift of finite type and let for

3118:   every subset $Y$ of $\OSS$, $\ideal(Y)$ be the 

3119:   ideal of $\csp{\OSS,\theta_{\OSS},\fg\al}$ generated

3120:   by the set

3121:   \begin{equation*}

3122:     \{\boomap[(\OSS,\theta_{\OSS},\fg\al)](A)\mid A\in

3123:     \Bx(\OSS,\theta_{\OSS},\fg\al),\ A\subseteq Y\}.

3124:   \end{equation*}  

3125:   Then the map

3126:   \begin{equation*}

3127:     Y\mapsto \ideal(Y)

3128:   \end{equation*}

3129:   is an injective order preserving (i.e., $ Y\subseteq Z\Rightarrow

3130:   \ideal(Y)\subseteq \ideal(Z)$) map from the set of open subsets $Y$ of

3131:   $\OSS$ with the property that $\osh(Y)\subseteq Y$ and 

3132:   $\osh\inv(Y)\subseteq Y$ to the set of ideals of $\O_{\OSS}$.

3133: \end{corollary}

3134: 

3135: 

3136: \subsection{Connections between the $\cs$-algebras of one- and

3137:   two-sided shift spaces}

3138: We will in this section let $\TSS$ be a two-sided shift space, and

3139: \begin{equation} \label{eq:twoone}

3140:   \OSS=\{(z_n)_{n\in \No}\mid (z_n)_{n\in \Z}\in \TSS\}

3141: \end{equation}

3142: be the corresponding one-sided shift space. We then have that map

3143: \begin{equation*}

3144:   \spro:(z_n)_{n\in \Z}\mapsto (z_n)_{n\in \No}

3145: \end{equation*}

3146: is a surjective continuous map from $\TSS$ to $\OSS$ and

3147: $\spro\circ\tsh=\osh\circ\spro$, where $\tsh$ and $\osh$ are the maps

3148: defined by \eqref{eq:twoshift} and \eqref{eq:oneshift}, respectively.

3149: 

3150: We will throughout this section let $(\TSS,\theta_\TSS,\fg{\al})$ be

3151: the discrete partial dynamical system associated to $(\TSS,\tsh)$ as

3152: done in Theorem \ref{kryds} and $(\OSS,\theta_{\OSS},\fg{\al})$ be the

3153: discrete partial dynamical system associated to $(\OSS,\osh)$ as done in

3154: Theorem \ref{theorem:one}. We let $(D_g^{\TSS})_{g\in\fg\al}$ and

3155: $(\theta_g^{\TSS})_{g\in\fg\al}$ denote the domains and partial one-to-one

3156: maps of $\theta_\TSS$, and $(D_g^{\OSS})_{g\in\fg\al}$ and

3157: $(\theta_g^{\OSS})_{g\in\fg\al}$ denote the domains and partial one-to-one

3158: maps of $\theta_{\OSS}$.

3159: 

3160: \begin{remark}

3161:   If $\TSS$ only contains finitely many elements, then every element of

3162:   $\TSS$ is periodic (meaning that there exists an $n\in\N$ such that

3163:   $\tsh^n(z)=z$), so $\spro$ is bijective, and since it also satisfies

3164:   that $\theta_g\circ\spro=\spro\circ\theta_g$ for every $g\in\fg\al$, it

3165:   easily follows from Theorem \ref{kryds} and \ref{theorem:one} that

3166:   there exists a $*$-isomorphism from $\O_{\OSS}$ to

3167:   $C(\TSS)\rtimes_{\tsh^\star}\Z$ which maps $\sum_{a\in\al}S_a$ to $U$

3168:   and $\ioss\big(1_{D_g^{\OSS}}\big)$ to $1_{D_g^{\TSS}}$  

3169:   for every $g\in\fg\al$, where $U$ is as in

3170:   Section \ref{sec:cross}, and $(S_a)_{a\in\al}$ and $\ioss$ are as in

3171:   Section \ref{sec:onesided}.

3172: \end{remark}

3173: 

3174: If $\TSS$ contains infinitely many elements, then

3175: $C(\TSS)\rtimes_{\tsh^\star}\Z$ and $\O_{\OSS}$ are, as we will see

3176: below, in general quit different. We will however show later that if

3177: $\TSS$ has a certain property, then $C(\TSS)\rtimes_{\tsh^\star}\Z$ is

3178: a quotient of $\O_{\OSS}$.

3179: 

3180: \begin{remark}

3181:   If $Y$ is an open subset of $\OSS$ with the property that

3182:   $\osh(Y)\subseteq Y$ and $\osh\inv(Y)\subseteq Y$, then

3183:   $\spro\inv(Y)$ is an open subset of $\TSS$ and

3184:   $\tsh(\spro\inv(Y))=\spro\inv(Y)$. This explains why

3185:   $C(\TSS)\rtimes_{\tsh^\star}\Z$ in general have more ideals than

3186:   $\O_{\OSS}$ if $\TSS$ (and thus $\OSS$) is of finite type.

3187: 

3188:   It is in fact easy to construct an example of a shift of finite type

3189:   $\TSS$ such that $\O_{\OSS}$ is simple, but $\TSS$ contains

3190:   infinitely many open $\tsh$-invariant subset and thus 

3191:   $C(\TSS)\rtimes_{\tsh^\star}\Z$ infinitely many ideals according to Section

3192:   \ref{subsec:idealkryds}. The full

3193:   two-shift $\{0,1\}^\Z$ will for example do the trick (in this case

3194:   $\O_{\OSS}$ will be the Cuntz-algebra $\O_2$, cf. \cite{MR0467330}).

3195: \end{remark}

3196: 

3197: If $\TSS$ and $\OSS$ are not of finite type, then it might happen

3198: that even though $\OSS$ is minimal (meaning that the only closed

3199: subsets $Y$ of $\OSS$ such that $\osh(Y)\subseteq Y$ are $\OSS$ and

3200: $\emptyset$, cf. \cite{MR97a:58050}*{\S13.7}), 

3201: there exists a subset $Y$ of $\OSS$ which is neither equal to $\OSS$

3202: nor $\emptyset$ and with the property that

3203: $\osh(Y)\subseteq Y$, 

3204: $\osh\inv(Y)\subseteq Y$ and $\forall x\in Y\exists

3205: (k,l)\in\ind:\ec{k}{x}{l}\subseteq Y$, and thus

3206: that $\O_{\OSS}$ is not simple. We will in Example \ref{exam:sub} see an

3207: example of this phenomenon.

3208: 

3209: We will now describe a class of shift spaces for which

3210: $C(\TSS)\rtimes_{\tsh^\star}\Z$ is a quotient of $\O_{\OSS}$, and we

3211: will then describe a subclass of this class for which we can show that

3212: the ideal of this quotient is a direct sum of a finite number of the

3213: compact operators $\K$

3214: 

3215: This result has been used in \cite{tmcseii} and \cite{tmcseiv} to

3216: compute the $K$-theory of $\O_{\OSS}$ and relate it to the $K$-theory of

3217: $C(\TSS)\rtimes_{\tsh^\star}\Z$ for these classes of shift spaces.

3218: 

3219: \begin{definition}

3220:   We say that a shift space $\TSS$ has

3221:   \emph{property $\coni$} if for every $\mu\in \laOSS$ there exists an

3222:   $x\in \OSS$ such that $\Past_{|\mu|}(x)=\{\mu\}$.

3223: \end{definition}

3224: 

3225: An element $z\in\TSS$ is called \emph{left special} if there exists

3226: $z'\in\TSS$ such that $\spro(z)=\spro(z')$, but $z_{-1}\ne z'_{-1}$.

3227: 

3228: Notice that $z\in\TSS$ is left special if and only if

3229: $\Past_1(\spro(z))$ consists of a least two elements.

3230: 

3231: \begin{definition}

3232:   We say that a shift space $\TSS$ has property $(**)$ if it has

3233:   property $(*)$ and the number of left special elements of $\TSS$

3234:   is finite, and no such left special word is periodic.

3235: \end{definition}

3236: 

3237: It has in \cite{tmcseii} been proved that every finite shift space and

3238: every minimal shift space with a finite number of left special

3239: elements (for example every shift space of a primitive substitution

3240: and every shift space of a Sturmian sequence) have property $(**)$, and

3241: that the shift space of a non-regular Toeplitz sequence has property

3242: $(*)$, but not necessarily property $(**)$.

3243: 

3244: We say that $z,z'\in \TSS$ are \emph{right shift tail equivalent} if

3245: there exist $m,N\in\Z$ such that $z_n=z'_{n+m}$ for all $n>N$. 

3246: 

3247: \begin{lemma} \label{lemma:faa}

3248:    Let $\TSS$ be a shift space which has property $(**)$ and let

3249:    $(k,l)\in\indd$. Then the set

3250:    \begin{equation*}

3251:      \{x\in\OSS\mid \Past_l(x_{[k,\infty[})\text{ contains more than

3252:        one element}\}

3253:    \end{equation*}

3254:    is finite, and every element of it is of the form $\spro(z)$, where

3255:    $z$ is an element of $\TSS$ which is right shift tail equivalent to

3256:    a left special element of $\TSS$. 

3257: \end{lemma}

3258: 

3259: \begin{proof}

3260:   Assume that $u,v,w\in\al^*$ are such that the last letter of

3261:   $u$ is different from the last letter of $v$ and

3262:   $uw,vw\in\Past_l(x_{[k,\infty[})$ ($w$ might be the empty

3263:   word). Then $\Past_1(wx_{[k,\infty[})$ consists of a least two

3264:   elements so $wx_{[k,\infty[}$ is equal to $\spro(z')$ for some left

3265:   special element $z'$. 

3266: 

3267:   Thus the set

3268:   \begin{equation*}

3269:      \{x\in\OSS\mid \Past_l(x_{[k,\infty[})\text{ contains more than

3270:        one element}\}

3271:    \end{equation*}

3272:    is finite, and every element of it is of the form $\spro(z)$, where

3273:    $z$ is an element of $\TSS$ which is right shift tail equivalent to

3274:    a left special element of $\TSS$. 

3275: \end{proof}

3276: 

3277: \begin{lemma} \label{lemma:nonperiodic}

3278:   Let $\TSS$ be a shift space which has property $(**)$, and let $z\in

3279:   X$ be right shift tail equivalent to a left special element. Then

3280:   $\spro(z)$ is not eventually periodic, meaning that there exist no

3281:   $k,N\in\N$ such that $z_n=z_{n+k}$ for all $n>N$. 

3282: \end{lemma}

3283: 

3284: \begin{proof}

3285:   Assume that $m,N_1\in\Z$ and $z_n=z'_{n+m}$ for all $n>N_1$ with $z'$

3286:   being a left special element, and that $k,N_2\in\N$ are such that

3287:   $z_{n}=z_{n+k}$ for all $n>N_2$. We then have that $z'_{n+m}=z'_{n+m+k}$ for

3288:   $n>\max\{N_1,N_2\}$, so the set $\{n\in\Z\mid z'_n\ne z'_{n+k}\}$

3289:   is bounded above, and since $z'$ is left special and thus not

3290:   periodic, this set is not empty. So we can define $n_0$ by the equation

3291:   \begin{equation*}

3292:     n_0=\max\{n\in\Z\mid z'_n\ne z'_{n+k}\}+1.

3293:   \end{equation*}

3294:   We then have that $z'_{n_0-1}\ne z'_{n_0-1+k}$ and that 

3295:   $z'_{n}= z'_{n+k}$ for $n\ge n_0$. 

3296: 

3297:   We define a sequence $z''$ by letting $z''_{n_0+i+lk}=z'_{n_0+i}$ for

3298:   $i\in\{0,1,\dotsc,k-1\}$ and $l\in\Z$. We then have for all $n\in\Z$ that 

3299:   $z''_{[n,\infty[}=z'_{[n+lk,\infty[}$, where $l\in\Z$ is chosen such that

3300:   $n+lk\ge n_0$, and thus that $z''\in\TSS$. We also have that $z''$ is

3301:   periodic, that $z''_{n_0-1}=z'_{n_0-1+k}\ne z'_{n_0-1}$ and that

3302:   $z''_{[n_0,\infty[}=z'_{[n_0,\infty[}$. Thus $\osh^{n_0}(z'')$ is an

3303:   element of $\TSS$ which is both periodic and left special, but that

3304:   contradicts the assumption that $\TSS$ has property $(**)$.

3305: \end{proof}

3306: 

3307: \begin{lemma} \label{lemma:iso}

3308:   Let $\TSS$ be a shift space which has property $(**)$. Then we have

3309:   that 

3310:   \begin{equation*}

3311:     \{\spro(z)\}\in\Bx(\OSS,\theta_{\OSS},\fg\al)

3312:   \end{equation*}

3313:   for every $z\in\TSS$ which is right shift equivalent to a left

3314:   special element. 

3315: \end{lemma}

3316: 

3317: \begin{proof}

3318:   Let $z$ be an element of $\TSS$ which is right shift equivalent to a left

3319:   special element. Then there exist a left special element $z'$ and

3320:   $m,N\in\Z$ such that $z_n=z'_{n+m}$ for all $n>N$. Since $z'$ is left

3321:   special, there exist $a,b\in\al$ such

3322:   that $a\ne b$ and $a,b\in\Past_1(\spro(z'))$. That means that 

3323:   \begin{equation*}

3324:     \spro(z')\in D_{a\inv}^{\OSS}\cap D_{b\inv}^{\OSS},

3325:   \end{equation*}

3326:   and since $\TSS$ only contains a finite number of special elements,

3327:   $D_{a\inv}^{\OSS}\cap D_{b\inv}^{\OSS}$ is finite. Thus there

3328:   exists a $k>0$ such that

3329:   \begin{equation*}

3330:     \{z'_{[k,\infty[}\}=D_{(au)\inv}^{\OSS}\cap D_{(bu)\inv}^{\OSS},

3331:   \end{equation*}

3332:   where $u=z'_{[0,k[}$. 

3333: 

3334:   Let us denote $\max\{m+N,0,m-k\}$ by $n_0$ and let $v=z'_{[k,k+n_0[}$  and

3335:   $w=z_{[0,n_0[}$. We then have that $\spro(z)=wy$ where

3336:   $y$ is the unique element in $\OSS$ such that

3337:   $vy=z'_{[k,\infty[}$. Thus it follows from Lemma \ref{lemma:onefund}

3338:   that 

3339:   \begin{equation*}

3340:     \{\spro(z)\}=\theta_{wv\inv}^{\OSS}\left(D_{(au)\inv}^{\OSS}\cap

3341:       D_{(bu)\inv }^{\OSS}\right) 

3342:     \in\Bx(\OSS,\theta_{\OSS},\fg\al). 

3343:   \end{equation*}

3344: \end{proof}

3345: 

3346: Remember that there exist an inclusion $\ioss$ of $\DX$ into

3347: $\O_{\OSS}$ (cf. Section \ref{sec:onesided})

3348: and a that $C(\TSS)$ sits inside

3349: $C(\TSS)\rtimes_{\tsh^\star}\Z$ (cf. Section \ref{sec:cross}). We will denote

3350: the inclusion of 

3351: $C(\TSS)$ into $C(\TSS)\rtimes_{\tsh^\star}\Z$ by $\crosmap$. We then

3352: have the follow result:

3353: 

3354: \begin{theorem} \label{theorem:toke}

3355:   Let $\TSS$ be a two-sided shift space which has property $\coni$

3356:   and let $\OSS$ be the corresponding one-sided shift space defined by

3357:   \eqref{eq:twoone}.

3358:   Then there are surjective $*$-homomorphisms $\kappa:\DX\to C(\TSS)$

3359:   and $\rho:\O_{\OSS}\to C(\TSS)\rtimes_{\tsh^\star}\Z$ making the

3360:   diagram

3361:   \begin{equation*}

3362:     \xymatrix{{\DX}\ar[d]_ - {\eta_{\mathcal{O}}}\ar[rr]^ -

3363:       {\kappa}&&{C(\TSS)}\ar[d]^ - {\eta_{\rtimes}}&\\

3364:       {\O_{\OSS}}\ar[rr]_-{\rho}&&{C(\TSS)\rtimes_{\tsh^\star}\Z}}

3365:   \end{equation*}

3366:   commute. We furthermore have that

3367:   \begin{equation*} 

3368:     \kappa(1_{\cy\mu\nu})=

3369:     \begin{cases}

3370:       1_{\left\{x\in\TSS\mid x_{[0,|v|[}=v\right\}}&\text{if }\exists

3371:       w\in\al^*:v=wu,\\ 

3372:       1_{\left\{x\in\TSS\mid x_{[|v|-|u|,|v|[}=u\right\}}&\text{if

3373:       }\exists w\in\al^*:u=wv,\\ 

3374:       0&\text{else,}

3375:     \end{cases}

3376:   \end{equation*}

3377:   for every $\mu,\nu\in\al^*$,

3378:   and that $\rho(S_a)=\eta_{\rtimes}(1_{D_a^{\TSS}})U$ for every

3379:   $a\in\al$ where $U$ 

3380:   is as in Section \ref{sec:cross} and $(S_a)_{a\in\al}$ and

3381:   $\cy\mu\nu$ are as in Section \ref{sec:onesided}.

3382: 

3383:   If $\TSS$ also has property $(**)$, then the kernel of $\rho$ is

3384:   isomorphic $\K^{\nTSS}$ where $\K$ is the $\cs$-algebra of compact

3385:   operator on an infinite dimensional separable Hilbert space and

3386:   $\nTSS$ is the number of right shift tail equivalence classes of $\TSS$

3387:   containing a left special element.

3388: \end{theorem}

3389: 

3390: \begin{proof} 

3391:   Let $(\TSS,\theta_\TSS,\fg{\al})$ be the discrete partial dynamical

3392:   system associated to $(\TSS,\tsh)$ as done in Theorem \ref{kryds}

3393:   and $(\OSS,\theta_{\OSS},\fg{\al})$ be the discrete partial

3394:   dynamical system associated to $(\OSS,\osh)$ as done in Theorem

3395:   \ref{theorem:one}. We let $(D_g^{\TSS})_{g\in\fg\al}$ and

3396:   $(\theta_g^{\TSS})_{g\in\fg\al}$ be the domains and partial

3397:   one-to-one maps of $\theta_\TSS$, and $(D_g^{\OSS})_{g\in\fg\al}$

3398:   and $(\theta_g^{\OSS})_{g\in\fg\al}$ the domains and partial

3399:   one-to-one maps of $\theta_{\OSS}$.

3400: 

3401:   Let for every $A\in \Bx(\OSS,\theta_{\OSS},\fg\al)$, $\psi(A)$ be

3402:   the subset of $\TSS$ defined by

3403:   \begin{equation*}

3404:     \psi(A)=\Bigl\{z\in \TSS\bigm| \forall (k,l)\in\ind\exists x\in

3405:     A:x_{[0,k[}=z_{[0,k[}\land

3406:     \Past_l(x_{[k,\infty[})=\{z_{[k-l,k[}\}\Bigr\}. 

3407:   \end{equation*}

3408:   We claim that the map $\psi$ is a Boolean homomorphism from

3409:   $\Bx(\OSS,\theta_{\OSS},\fg\al)$ to $\Bx(\TSS,\theta_\TSS,\fg\al)$

3410:   which satisfies that

3411:   \begin{equation*}

3412:     \psi\bigl(\theta_g^{\OSS}(A)\bigr)=\theta_g^{\TSS}\bigl(\psi(A)\bigr)

3413:   \end{equation*}

3414:   for all $A\in \Bx(\OSS,\theta_{\OSS},\fg\al)$ and $g\in \fg\al$.

3415: 

3416:   We will prove that by establishing a sequence of claims.

3417: 

3418:   \begin{claim}

3419:     The map $\psi$ is a Boolean homomorphism.

3420:   \end{claim}

3421: 

3422:   \begin{proof}

3423:     We will prove that $\psi$ is a Boolean homomorphism by showing

3424:     that $\psi(A\cap B)=\psi(A)\cap \psi(B)$ and $\psi(\OSS\setminus

3425:     A)=\TSS\setminus \psi(A)$ for all

3426:     $A,B\in\Bx(\OSS,\theta_{\OSS},\fg\al)$.

3427:   

3428:     Let $A,B\in \Bx(\OSS,\theta_{\OSS},\fg\al)$. It is obvious that

3429:     $\psi(A\cap B)\subseteq \psi(A)\cap \psi(B)$.  Assume that $z\in

3430:     \psi(A)\cap \psi(B)$. There exist by Lemma \ref{lemma:clopen}

3431:     $(k_a,l_a),(k_b,l_b)\in \indd$ such that the two implications

3432:     \begin{equation*}

3433:       x\in A\Rightarrow \ec{k_a}{x}{l_a}\subseteq A

3434:     \end{equation*}

3435:     and

3436:     \begin{equation*}

3437:       x\in B\Rightarrow \ec{k_b}{x}{l_b}\subseteq B,

3438:     \end{equation*}

3439:     hold for all $x\in\OSS$.

3440:   

3441:     Let $(k_0,l_0)\in\indd$ and choose $(k,l)\in\indd$ such that we

3442:     have that

3443:     \begin{equation*}

3444:       (k,l)\gee (k_a,l_a),(k_b,l_b), (k_0,l_0),

3445:     \end{equation*}

3446:     and choose $x^A\in A$ and $x^B\in B$ such that we have that

3447:     \begin{equation*}

3448:       z_{[0,k[}=x^A_{[0,k[}=x^B_{[0,k[} \text{ and }\Past_l(x^A_{[k,\infty[})=

3449:       \Past_l(x^B_{[k,\infty[})=\{z_{[k-l,k[}\}.

3450:     \end{equation*}

3451:     Then $x^A\eq x^B$, so $x^A\in A\cap B$, and since

3452:     $x^A_{[0,k_0[}=z_{[0,k_0[}$ and

3453:     $\Past_{l_0}((x^A_{[k_0,\infty[})=\{z_{[k_0-l_0,k[}\}$, this shows

3454:     that $z\in \psi(A\cap B)$. Thus $\psi(A\cap B)=\psi(A)\cap

3455:     \psi(B)$.

3456:   

3457:     Assume now that $z\in \psi(\OSS\setminus A)$. We then have that if

3458:     $x\in\OSS$, $z_{[0,k_a[}=x_{[0,k_a[}$ and

3459:     $\Past_{l_a}(x_{[k_a,\infty[}) = \{z_{[k_a-l_a,k_a[}\}$, then

3460:     $x\in\OSS\setminus A$. Thus $z\in\TSS\setminus\psi(A)$.

3461:   

3462:     If $z\in\TSS\setminus\psi(A)$, then there is a $(k_z,l_z)\in\indd$

3463:     such that we for every $x\in A$ have that

3464:     \begin{equation*}

3465:       z_{[0,k_z[}\ne x_{[0,k_z[}\text{ or

3466:       }\Past_{l_z}(x_{[k_z,\infty[})\ne \{z_{[k_z-l_z,k_z[}\}. 

3467:     \end{equation*}

3468:     Let $(k_0,l_0)\in\indd$ and choose $(k,l)\gee

3469:     (k_z,l_z),(k_0.l_0)$. Since $\TSS$ has property $\coni$, there

3470:     exists an $x\in \OSS$ such that $z_{[0,k[}=x_{[0,k[}$ and

3471:     $\Past_l(x_{[k,\infty[}) = \{z_{[k-l,k[}\}$. This $x$ must belong

3472:     to $\OSS\setminus A$, and since $z_{[0,k_0[}=x_{[0,k_0[}$ and

3473:     $\Past_{l_0}(x_{[k_0,\infty[}) = \{z_{[k_0-l_0,k_0[}\}$, this

3474:     shows that $z\in \psi(\OSS\setminus A)$. Thus $\psi(\OSS\setminus

3475:     A)=\TSS\setminus\psi(A)$.

3476:   \end{proof}

3477: 

3478:   \begin{claim} \label{et} We have that

3479:     $\psi\bigl(\theta_a^{\OSS}(A)\bigr)=\theta_a^{\TSS}\bigl(\psi(A)\bigr)$

3480:     for every $A\in \Bx(\OSS,\theta_{\OSS},\fg\al)$ and every

3481:     $a\in\al$.

3482:   \end{claim}

3483: 

3484:   \begin{proof}

3485:     Let $z\in \psi\bigl(\theta_a^{\OSS}(A)\bigr)$ and let

3486:     $(k,l)\in\indd$. Then there is an $x\in\theta_a^{\OSS}(A)$ such

3487:     that $x_{[0,k+1[}=z_{[0,k+1[}$ and

3488:     $\Past_l(x_{[k+1,\infty[})=\{z_{[k+1-l,k+1[}\}$. That means that

3489:     there is a $y\in A\cap D_{a\inv}^{\OSS}$ such that

3490:     $x=\theta_a^{\OSS}(y)$. We then have that $ z_0=x_0=a$, that

3491:     $y_{[0,k[}=x_{[1,k+1[}=z_{[1,k+1[}$ and that

3492:     \begin{equation*}

3493:       \Past_l(y_{[k,\infty[})=\Past_l(x_{[k+1,\infty[})=\{z_{[k+1-l,k+1[}\}.

3494:     \end{equation*}

3495:     Thus $\tsh(z)\in \psi(A)$, and $z_0=a$, which shows that $z\in

3496:     \theta_a^{\TSS}\bigl(\psi(A)\bigr)$.

3497:   

3498:     Now let $z\in \theta_a^{\OSS}\bigl(\psi(A)\bigr)$. Then there is a

3499:     $z'\in \psi(A)$ such that $z=\theta_a^{\TSS}(z')$. That means that

3500:     there for every $(k,l)\in\indd$, is an $x\in A$ such that

3501:     \begin{equation*}

3502:       z'_{[0,k[}= x_{[0,k[}\text{ and }\Past_l(x_{[k,\infty[})=

3503:       \{z'_{[k-l,k[}\}. 

3504:     \end{equation*}

3505:     We especially have that $\Past_1(x)=\{z'_{-1}\}=\{a\}$, so $x\in

3506:     D_{a\inv}^{\OSS}$, and we furthermore have that

3507:     \begin{equation*}

3508:       z_{[0,k+1[}=az'_{[0,k[}=ax_{[0,k[}=\bigl(\theta_a^{\OSS}(x)\bigr)_{[0,k+1[}

3509:     \end{equation*}

3510:     and that

3511:     \begin{equation*}

3512:       \Past_l\bigl((\theta_a^{\OSS}(x)\bigr)_{[k+1,\infty[})=

3513:       \Past_l(x_{[k,\infty[})= \{z'_{[k-l,k[}\}= \{z_{[k+1-l,k+1[}\},

3514:     \end{equation*}

3515:     which shows that $z\in\psi(\theta_a^{\OSS}(A))$.

3516:   \end{proof}

3517:   

3518:   \begin{claim} \label{to} We have that

3519:     $\psi\bigl(\theta_{a\inv}^{\OSS}(A)\bigr)=

3520:     \theta_{a\inv}^{\TSS}\bigl(\psi(A)\bigr)$ for every $a\in\al$ and

3521:     every $A\in \Bx(\OSS,\theta_{\OSS},\fg\al)$.

3522:   \end{claim}

3523:   

3524:   \begin{proof}

3525:     Let $z\in \psi\bigl(\theta_{a\inv}^{\OSS}(A)\bigr)$ and let

3526:     $(k,l)\in\indd$. Then there is an $x\in\theta_{a\inv}^{\OSS}(A)$

3527:     such that $x_{[0,k+1[}=z_{[0,k+1[}$ and

3528:     $\Past_{l+k+2}(x_{[k+1,\infty[})=\{z_{[-l-1,k+1[}\}$. That means

3529:     that there is a $y\in A\cap D_{a}^{\OSS}$ such that

3530:     $x=\theta_{a\inv}^{\OSS}(y)$, and we then have that

3531:     \begin{equation*}

3532:       a=y_0\in\Past_1(x)=\{z_{-1}\},

3533:     \end{equation*}

3534:     and thus that

3535:     \begin{equation*}

3536:       y_{[0,k[}=ax_{[0,k-1[}=z_{[-1,k-1[}\text{ and }

3537:       \Past_l(y_{[k,\infty[})=\Past_l(x_{[k-1,\infty[})=\{z_{[k-1-l,k-1[}\}

3538:     \end{equation*}

3539:     if $k>0$, and

3540:     \begin{equation*}

3541:       \Past_l(y_{[k,\infty[})=\Past_l(ax)=\{z_{[k-1-l,k-1[}\}

3542:     \end{equation*}

3543:     if $k=0$.  Thus $\tsh\inv(z)\in \psi(A)$ and $z_{-1}=a$, which shows

3544:     that

3545:     \begin{equation*}

3546:       z=\theta_{a\inv}^{\TSS}\bigl(\tsh(z)\bigr)

3547:       \in\theta_{a\inv}^{\TSS}\bigl(\psi(A)\bigr).

3548:     \end{equation*}

3549:   

3550:     Now let $z\in \theta_{a\inv}^{\TSS}\bigl(\psi(A)\bigr)$. We then

3551:     have that there is a $z'\in \psi(A)\cap D_a^{\TSS}$ such that

3552:     $z=\theta_{a\inv}^{\TSS}(z')$. That means that there for every

3553:     $(k,l)\in\indd$ is an $x\in A$ such that 

3554:     $z'_{[0,k+1[}=x_{[0,k+1[}$ and 

3555:     $\Past_l(x_{[k+1,\infty[})=\{z'_{[k+1-l,k+1[}\}$. 

3556:     We especially have that $x_0=z'_0=a$, so

3557:     $x\in A\cap D_a$, and we also have that

3558:     \begin{equation*}

3559:       z_{[0,k[}=z'_{[1,k+1[}=x_{[1,k+1[}=\bigl(\theta_{a\inv}^{\OSS}(x)\bigr)_{[0,k[}

3560:     \end{equation*}

3561:     and that

3562:     \begin{equation*}

3563:       \Past_l\Bigl(\bigl(\theta_{a\inv}^{\OSS}(x)\bigr)_{[k,\infty[}\Bigr)=

3564:       \Past_l(x_{[k+1,\infty[})= 

3565:       \{z'_{[k+1-l,k+1[}\}=\{z_{[k-l,l[}\}.

3566:     \end{equation*}

3567:     This shows that $z\in\psi\bigl(\theta_{a\inv}^{\OSS}(A)\bigr)$.

3568:   \end{proof}

3569:   

3570:   It follows from Claim \ref{et} and Claim \ref{to} and the definition

3571:   of $\theta_g^{\OSS}$ and $\theta_g^{\TSS}$ that

3572:   \begin{equation*}

3573:     \psi\bigl(\theta_g^{\OSS}(A)\bigr)=\theta_g^{\TSS}\bigl(\psi(A)\bigr)

3574:   \end{equation*}

3575:   for every $A\in\Bx(\OSS,\theta_{\OSS},\fg\al)$ and every

3576:   $g\in\fg\al$.

3577: 

3578:   Thus if we let $(s_g^{\OSS})_{g\in\fg\al}$ denote the generators of

3579:   $\csp{\OSS,\theta_{\OSS},\fg\al}$ and $(s_g^{\TSS})_{g\in\fg\al}$

3580:   the generators of $\csp{\TSS,\theta_{\TSS},\fg\al}$, then it follows

3581:   from Theorem \ref{theorem:surmap} that there exists a surjective

3582:   $*$-homomorphism $\phi$ from $\csp{\OSS,\theta_{\OSS},\fg\al}$ to

3583:   $\csp{\TSS,\theta_\TSS,\fg\al}$ which maps $s_g^{\OSS}$ to

3584:   $s_g^{\TSS}$ for every $g\in\fg\al$, and thus from Theorem

3585:   \ref{kryds} and \ref{theorem:one} that there is a surjective

3586:   $*$-homomorphism $\rho:\O_{\OSS}\to C(\TSS)\rtimes_{\tsh^\star}\Z$

3587:   which maps $\sum_{a\in\al}S_a$ to $U$ and $\ioss(1_{D_g})$ to

3588:   $\eta_{\rtimes}(1_{D_g})$ for every $g\in\fg\al$. We then have that

3589:   \begin{equation*}

3590:     \rho(S_a)=\rho\left(\ioss\biggl(1_{D_a}\sum_{a'\in\al}S_{a'}\biggr)\right)=

3591:     \eta_{\rtimes}(1_{D_a})U

3592:   \end{equation*}

3593:   for every $a\in\al$, and that

3594:   \begin{equation*}

3595:     \begin{split}

3596:       \rho\bigl(\ioss(1_{\cy\mu\nu})\bigr)&=

3597:       \rho\bigl(\ioss(1_{D_\nu})\bigr)

3598:       \rho\bigl(\ioss(1_{D_{\nu\mu\inv}})\bigr)\\

3599:       &= \eta_{\rtimes}(1_{D_\nu})\eta_{\rtimes}(1_{D_{\nu\mu\inv}})\\

3600:       &= \begin{cases} \eta_{\rtimes}\Bigl(1_{\left\{x\in\TSS\mid

3601:             x_{[0,|v|[}=v\right\}}\Bigr)&

3602:         \text{if }\exists w\in\al^*:v=wu,\\

3603:         \eta_{\rtimes}\Bigl(1_{\left\{x\in\TSS\mid

3604:             x_{[|v|-|u|,|v|[}=u\right\}}\Bigr)& \text{if }\exists w\in\al^*:u=wv,\\

3605:         0&\text{else,}

3606:       \end{cases}

3607:     \end{split}

3608:   \end{equation*}

3609:   for $u,v\in\al^*$, according to Lemma \ref{lemma:onefund} and

3610:   \ref{lemma:twofund}.  Since $\Dx$ is generated by

3611:   $\{1_{\cy\mu\nu}\mid u,v\in\al^*\}$ and $C(X)$ is generated by the

3612:   set

3613:   \begin{equation*}

3614:     \Bigl\{1_{\{x\in \TSS\mid x_{[k-|u|,k[}=u\}}\bigm| u\in\al^*,\

3615:     k\in\Z\Bigr\},

3616:   \end{equation*}

3617:   it follows that there exists a surjective $*$-homomorphism

3618:   $\kappa:\DX\to C(\TSS)$ with the desired properties.

3619: 

3620:   Assume now that $\TSS$ has property $(**)$. Let $\rte$ be the set of

3621:   right shift tail equivalence classes of $\TSS$ which contains a left

3622:   special element. We will show that the kernel of $\phi$, and thus

3623:   the kernel of $\rho$, is isomorphic to $\nTSS$ copies of $\K$ by

3624:   constructing a family

3625:   \begin{equation*}

3626:     (e^\jj_{x,y})_{\bigl(\jj\in\rte,\ x,y\in\spro(\jj)\bigr)}

3627:   \end{equation*}

3628:   of non-zero elements of $\csp{\OSS,\theta_{\OSS},\fg\al}$ such that

3629:   the equations

3630:   \begin{equation*}

3631:     (e^\jj_{x,y})^*=e^\jj_{y,x}\text{ and }

3632:     e^\jj_{x,y}e^{\jj'}_{x',y'}=

3633:     \begin{cases}

3634:       e^\jj_{x,y'}&\text{if }\jj=\jj'\text{ and }y=x',\\

3635:       0&\text{else,}

3636:     \end{cases}

3637:   \end{equation*}

3638:   hold for all $\jj,\jj'\in\rte$ and all $x,y\in\spro(\jj)$ and

3639:   $x',y'\in\spro(\jj')$, and such that we have that

3640:   \begin{equation*}

3641:     \spc\bigr\{e^\jj_{x,y}\bigm|\jj\in\rte,\ x,y\in\spro(\jj)\bigr\}=\ker\phi.

3642:   \end{equation*}

3643: 

3644:   So let $\jj\in\rte$ and $x,y\in\spro(\jj)$. Then there exist

3645:   $n,m\in\N$ such that $x_{[n,\infty[}=y_{[m,\infty[}$. It follows

3646:   from Lemma \ref{lemma:iso} that $\{x\}$ and $\{y\}$ belong to

3647:   $\Bx(\OSS,\theta_{\OSS},\fg\al)$ so we can define an element

3648:   $e^\jj_{x,y}$ in $\csp{\OSS,\theta_{\OSS},\fg\al}$ by the equation

3649:   \begin{equation*}

3650:     e^\jj_{x,y}= \boomap[(\OSS,\theta_{\OSS},\fg\al)]\bigl(\{x\}\bigr)\

3651:     s_{x_{[0,n[}y\inv_{[0,m[}}^{\OSS}\

3652:     \boomap[(\OSS,\theta_{\OSS},\fg\al)] \bigl(\{y\}\bigr). 

3653:   \end{equation*}

3654:   Notice that $e^\jj_{x,y}$ does not depend on the choice of $n$ and

3655:   $m$, because if $k,l\in\N$ also satisfy that

3656:   $x_{[k,\infty[}=y_{[l,\infty[}$, then we have that

3657:   \begin{equation*}

3658:     x_{n+l+i}=y_{m+l+i}=x_{m+k+i}

3659:   \end{equation*}

3660:   for all $i\in\N$, and since $x$ is not eventually periodic according

3661:   to Lemma \ref{lemma:nonperiodic} that means that $n+l=m+k$ and thus

3662:   that

3663:   \begin{equation*}

3664:     x_{[0,n[}y\inv_{[0,m[}= x_{[0,n+l[}y\inv_{[0,m+l[}=

3665:     x_{[0,m+k[}y\inv_{[0,m+l[}= x_{[0,k[}y\inv_{[0,l[}. 

3666:   \end{equation*}

3667: 

3668:   It is clear that $(e^\jj_{x,y})^*=e^\jj_{y,x}$ and that

3669:   $e^\jj_{x,y}e^{\jj'}_{x',y'}=0$ if $y\ne x'$. Assume now that

3670:   $\jj\in\rte$, that $x,y,z\in\spro(\jj)$ and that $k,l,m,n\in\N$ are

3671:   such that $x_{[n,\infty[}=y_{[m,\infty[}$ and

3672:   $y_{[k,\infty[}=z_{[l,\infty[}$. We then have that

3673:   \begin{equation*}

3674:     y\in D_{\bigl(x_{[0,n[}y\inv_{[0,m[}\bigr)\inv}^{\OSS}\text{ and }

3675:     \theta_{x_{[0,n[}y\inv_{[0,m[}}^{\OSS}(y)=x.

3676:   \end{equation*} 

3677:   Thus it follows from Lemma \ref{booinvariant} that we have that

3678:   \begin{equation*}

3679:     \begin{split}

3680:       s_{x_{[0,n[}y\inv_{[0,m[}}^{\OSS}\

3681:       \boomap[(\OSS,\theta_{\OSS},\fg\al)]\bigl(\{y\}\bigr)&=

3682:       s_{x_{[0,n[}y\inv_{[0,m[}}^{\OSS}

3683:       \Bigl(s^{\OSS}_{x_{[0,n[}y\inv_{[0,m[}}\Bigr)^*

3684:       s_{x_{[0,n[}y\inv_{[0,m[}}^{\OSS}\

3685:       \boomap[(\OSS,\theta_{\OSS},\fg\al)]\bigl(\{y\}\bigr)\\

3686:       &= s_{x_{[0,n[}y\inv_{[0,m[}}^{\OSS}\

3687:       \boomap[(\OSS,\theta_{\OSS},\fg\al)]\bigl(\{y\}\bigr)\

3688:       \Bigl(s_{x_{[0,n[}y\inv_{[0,m[}}^{\OSS}\Bigr)^*

3689:       s_{x_{[0,n[}y\inv_{[0,m[}}^{\OSS}\\

3690:       &=\boomap[(\OSS,\theta_{\OSS},\fg\al)]\bigl(\{x\}\bigr)\

3691:       s_{x_{[0,n[}y\inv_{[0,m[}}^{\OSS}.

3692:     \end{split}

3693:   \end{equation*}

3694:   We also have that $x\in D_{x_{[0,n[}y\inv_{[0,m[}}^{\OSS}$, and thus

3695:   that

3696:   \begin{equation*}

3697:     \begin{split}

3698:       &\boomap[(\OSS,\theta_{\OSS},\fg\al)]\bigl(\{x\}\bigr)\

3699:       s_{x_{[0,n[}y\inv_{[0,m[}}^{\OSS} s_{y_{[0,k[}z\inv_{[0,l[}}^{\OSS} \\

3700:       &\qquad =\boomap[(\OSS,\theta_{\OSS},\fg\al)]\bigl(\{x\}\bigr)\

3701:       \boomap[(\OSS,\theta_{\OSS},\fg\al)]

3702:       \Bigl(D_{x_{[0,n[}y\inv_{[0,m[}}^{\OSS}\Bigr)\

3703:       s_{x_{[0,n[}y\inv_{[0,m[}y_{[0,k[}z\inv_{[0,l[}}^{\OSS}\\

3704:       &\qquad = \boomap[(\OSS,\theta_{\OSS},\fg\al)]\bigl(\{x\}\bigr)\

3705:       s_{x_{[0,n+k[ z\inv_{[0,m+l[}}}^{\OSS}

3706:     \end{split}

3707:   \end{equation*}

3708:   according to (\ref{eq:oxc}) and (\ref{eq:oxd}). Thus we have that

3709:   \begin{equation*}

3710:     \begin{split}

3711:       e^\jj_{x,y}e^\jj_{y,z}&=

3712:       \begin{multlined}[t]

3713:         \boomap[(\OSS,\theta_{\OSS},\fg\al)]\bigl(\{x\}\bigr)\

3714:         s_{x_{[0,n[}y\inv_{[0,m[}}^{\OSS}\

3715:         \boomap[(\OSS,\theta_{\OSS},\fg\al)]\bigl(\{y\}\bigr)\\

3716:         \boomap[(\OSS,\theta_{\OSS},\fg\al)]\bigl(\{y\}\bigr)\

3717:         s_{y_{[0,k[}z\inv_{[0,l[}}^{\OSS}\

3718:         \boomap[(\OSS,\theta_{\OSS},\fg\al)]\bigl(\{z\}\bigr)

3719:       \end{multlined}\\

3720:       &= \boomap[(\OSS,\theta_{\OSS},\fg\al)]\bigl(\{x\}\bigr)\

3721:       s_{x_{[0,n[}y\inv_{[0,m[}}^{\OSS}

3722:       s_{y_{[0,k[}z\inv_{[0,l[}}^{\OSS}\

3723:       \boomap[(\OSS,\theta_{\OSS},\fg\al)]\bigl(\{z\}\bigr)\\

3724:       &= \boomap[(\OSS,\theta_{\OSS},\fg\al)]\bigl(\{x\}\bigr)\

3725:       s_{x_{[0,n+k[}z\inv_{[0,m+l[}}^{\OSS}\

3726:       \boomap[(\OSS,\theta_{\OSS},\fg\al)]\bigl(\{z\}\bigr)\\

3727:       &=e^\jj_{x,z}.

3728:     \end{split}

3729:   \end{equation*}

3730:   Finally we notice that it follows from Corollary \ref{boomapinj}

3731:   that $e^\jj_{x,y}$ is non-zero, because we have that

3732:   \begin{equation*}

3733:     e^\jj_{x,y}\bigl(e^\jj_{x,y}\bigr)^*=e^\jj_{x,x}=

3734:     \boomap[(\OSS,\theta_{\OSS},\fg\al)]\bigl(\{x\}\bigr). 

3735:   \end{equation*}

3736: 

3737:   It follows from Theorem \ref{theorem:surmap} that the kernel of

3738:   $\phi$ is generated by the set

3739:   \begin{equation*}

3740:     \{\boomap[\OSS,\theta_{\OSS},\fg\al](A)\mid \psi(A)=\emptyset\},

3741:   \end{equation*}

3742:   and since

3743:   $\psi\Bigl(\boomap[(\OSS,\theta_{\OSS},\fg\al)]\bigl(\{x\}\bigr)\Bigr)=

3744:   \emptyset$ if $x\in\spro(\jj)$ for some $\jj\in\rte$, we have that

3745:   the set

3746:   \begin{equation*}

3747:     \spc\bigl\{e^\jj_{x,y}\mid

3748:     \jj\in\rte,\ x,y\in\spro(\jj)\bigr\}

3749:   \end{equation*}

3750:   is contained in the kernel of $\phi$.

3751: 

3752:   Assume that $A\in\Bx(\OSS,\theta_{\OSS},\fg\al)$ and that

3753:   $\psi(A)=\emptyset$. Choose by Lemma \ref{lemma:clopen}

3754:   $(k,l)\in\indd$ such that the implication

3755:   \begin{equation*}

3756:     x\in A\Rightarrow \ec{k}{x}{l}\subseteq A

3757:   \end{equation*}

3758:   holds for all $x\in\OSS$. Then we have that

3759:   \begin{equation*}

3760:     A=\bigcup_{x\in A}\ec{k}{x}{l}.

3761:   \end{equation*}

3762:   Assume that $x\in A$ and that the number of elements of

3763:   $\Past_l(x_{[k,\infty[})$ is one. Then it follows from Lemma

3764:   \ref{lemma:twofund} that

3765:   \begin{equation*}

3766:     \begin{split}

3767:       \psi(\ec{k}{x}{l}) &=

3768:       \psi\left(\theta_{x_{[0,k[}}^{\OSS}\Biggl(D_{\nu\inv}^{\OSS}\cap

3769:         \biggl(

3770:         \smashoperator[r]{\bigcap_{\mu\in\neg\Past_l(\osh^k(x))}}

3771:         \OSS\setminus D_{\mu\inv}^{\OSS}\biggr)\Biggr)\right)\\

3772:       &= \theta_{x_{[0,k[}}^{\TSS}\left(D_{\nu\inv}^{\TSS}\cap \Biggl(

3773:         \smashoperator[r]{\bigcap_{\mu\in\neg\Past_l(\osh^k(x))}}

3774:         \OSS\setminus D_{\mu\inv}^{\TSS}\Biggr)\right)\\

3775:       &= \theta_{x_{[0,k[}}^{\TSS}\bigl(D_{\nu\inv}^{\TSS}\bigr)\ne

3776:       \emptyset,

3777:     \end{split}

3778:   \end{equation*}

3779:   where $\nu$ is the unique element of $\Past_l(x_{[k,\infty[})$ and

3780:   $\neg\Past_l(x_{[k,\infty[})= \bigl\{\mu\in\laOSS\mid |\mu|=l,\

3781:   \mu\notin\Past_l(x_{[k,\infty[})\bigr\}$, but this contradict our

3782:   assumption that $\psi(A)=\emptyset$.

3783: 

3784:   Thus $\Past_l(x_{[k,\infty[})$ consists of at least two elements for

3785:   every $x\in A$, so it follows from Lemma \ref{lemma:faa} that $A$ is

3786:   finite and that every element of $A$ is of the form $\spro(z)$ for

3787:   some $z\in\TSS$ which is right shift tail equivalent to a left

3788:   special element. Hence we get by Lemma \ref{lemma:iso} that

3789:   \begin{multline*}

3790:     \boomap[(\OSS,\theta_{\OSS},\fg\al)](A)= 

3791:     \sum_{x\in A}\boomap[(\OSS,\theta_{\OSS},\fg\al)](\{x\})=\\ 

3792:     \sum_{x\in A}e^\xx_{x,x}\in \spc\bigl\{e^\jj_{x,y}\bigm|\jj\in\rte,\

3793:     x,y\in\spro(\jj)\bigr\},

3794:   \end{multline*}

3795:   where for every $x\in A$, $\xx$ denotes the right shift tail

3796:   equivalence class which contains an element $z$ such that

3797:   $\spro(z)=x$.

3798: 

3799:   So to show that the equality

3800:   \begin{equation*}

3801:     \spc\bigl\{e^\jj_{x,y}\bigm|\jj\in\rte,\ x,y\in\spro(\jj)\bigr\}=\ker\phi,

3802:   \end{equation*}

3803:   holds, we just have to show that the subset

3804:   \begin{equation*}

3805:     \spc\bigl\{e^\jj_{x,y}\bigm|\jj\in\rte,\ x,y\in\spro(\jj)\bigr\}

3806:   \end{equation*}

3807:   is an ideal of $\csp{\OSS,\theta_{\OSS},\fg\al}$, and since

3808:   $\csp{\OSS,\theta_{\OSS},\fg\al}$ is generated by the set

3809:   $\bigl\{s_g^{\OSS}\bigm| g\in\fg\al\bigr\}$, it is enough to prove

3810:   that $s_g^{\OSS}e^\jj_{x,y}$ and $e^\jj_{x,y}s_g^{\OSS}$ belong to

3811:   $\spc\bigl\{e^\jj_{x',y'}\bigm|\jj\in\rte,\

3812:   x',y'\in\spro(\jj)\bigr\}$ for every $g\in\fg\al$, $\jj\in\rte$ and

3813:   $x,y\in\spro(\jj)$.

3814: 

3815:   So let $g\in\fg\al$, $\jj\in\rte$ and $x,y\in\spro(\jj)$, and let

3816:   $n,m\in\N$ such that $x_{[n,\infty[}=y_{[m,\infty[}$. It follows

3817:   from Lemma \ref{lemma:onestruc} that if $s_g^{\OSS}\ne 0$, then

3818:   there exist $\mu,\nu\in\al^*$ such that $g=\mu\nu\inv$. If $y$ does

3819:   not belong to $D_{\mu\nu\inv}^{\OSS}$ then we have that

3820:   \begin{equation*}

3821:     \boomap[(\OSS,\theta_{\OSS},\fg\al)]\bigl(\{y\}\bigr)\ s_{\mu\nu\inv}^{\OSS} =

3822:     \boomap[(\OSS,\theta_{\OSS},\fg\al)]\bigl(\{y\}\bigr)\

3823:     \boomap[(\OSS,\theta_{\OSS},\fg\al)]\bigl(D_{\mu\nu\inv}^{\OSS}\bigr)\

3824:     s_{\mu\nu\inv}^{\OSS}

3825:     =\emptyset,

3826:   \end{equation*}

3827:   from which it follows that

3828:   \begin{equation*}

3829:     e^\jj_{x,y} s_g^{\OSS}= \boomap[(\OSS,\theta_{\OSS},\fg\al)]\bigl(\{x\}\bigr)\

3830:     s_{x_{[0,n[}y\inv_{[0,m[}}^{\OSS}\

3831:     \boomap[(\OSS,\theta_{\OSS},\fg\al)]\bigl(\{y\}\bigr)\ 

3832:     s_{\mu\nu\inv}^{\OSS} =0. 

3833:   \end{equation*}

3834:   If $y\in D_{\mu\nu\inv}^{\OSS}$, then we have according to Lemma

3835:   \ref{lemma:onefund} and \ref{booinvariant} that

3836:   $\theta_{\nu\mu\inv}^{\OSS}(y)\in\spro(\jj)$,

3837:   $\theta_{\nu\mu\inv}^{\OSS}(y)_{[m+|\nu|,\infty[}=

3838:   x_{[n+|\mu|,\infty[}$ and that

3839:   \begin{equation*}

3840:     \begin{split}

3841:       \boomap[(\OSS,\theta_{\OSS},\fg\al)]\bigl(\{y\}\bigr)\

3842:       s_{\mu\nu\inv}^{\OSS} &=

3843:       \boomap[(\OSS,\theta_{\OSS},\fg\al)]\bigl(\{y\}\bigr)\

3844:       s_{\mu\nu\inv}^{\OSS}

3845:       \bigl(s_{\mu\nu\inv}^{\OSS}\bigr)^*s_{\mu\nu\inv}^{\OSS}\\

3846:       &= s_{\mu\nu\inv}^{\OSS} \bigl(s_{\mu\nu\inv}^{\OSS}\bigr)^*

3847:       \boomap[(\OSS,\theta_{\OSS},\fg\al)]\bigl(\{y\}\bigr)\

3848:       s_{\mu\nu\inv}^{\OSS}\\

3849:       &= s_{\mu\nu\inv}^{\OSS}\ \boomap[(\OSS,\theta_{\OSS},\fg\al)]

3850:       \bigl(\{\theta_{\nu\mu\inv}^{\OSS}(y)\}\bigr).

3851:     \end{split}

3852:   \end{equation*}

3853:   We further more have that

3854:   \begin{equation*}

3855:     \begin{split}

3856:       y\inv_{[0,m+|\mu|[}uv\inv &= \bigl(uy_{[|\mu|,m+|\mu|[}\bigr)\inv uv\inv\\

3857:       &= y_{[|\mu|,m+|\mu|[}\inv v\inv\\

3858:       &= \left(\theta_{vu\inv}^{\OSS}(y)_{[0,m+|\nu|[}\right)\inv,

3859:     \end{split}

3860:   \end{equation*}

3861:   and hence that

3862:   \begin{multline*}

3863:     s_{x_{[0,n+|\mu|[}y\inv_{[0,m+|\mu|[}}^{\OSS} s_{\mu\nu\inv}^{\OSS} =\\

3864:     \boomap[(\OSS,\theta_{\OSS},\fg\al)]

3865:     \Bigl(D_{x_{[0,n+|\mu|[}y\inv_{[0,m+|\mu|[}}^{\OSS}\Bigr)

3866:     s_{x_{[0,n+|\mu|[}\bigl(\theta_{vu\inv}^{\OSS}(y)_{[0,m+|\nu|[}\bigr)\inv}^{\OSS}

3867:   \end{multline*}

3868:   according to \eqref{eq:oxc} and \eqref{eq:oxd} of Definition

3869:   \ref{def:ox}.

3870: 

3871:   Thus we have that

3872:   \begin{equation*}

3873:     \begin{split}

3874:       e^\jj_{x,y} s_g^{\OSS} &=

3875:       \boomap[(\OSS,\theta_{\OSS},\fg\al)]\bigl(\{x\}\bigr)\

3876:       s_{x_{[0,n[}y\inv_{[0,m[}}^{\OSS}\

3877:       \boomap[(\OSS,\theta_{\OSS},\fg\al)]\bigl(\{y\}\bigr)\

3878:       s_{\mu\nu\inv}^{\OSS}\\

3879:       &= \boomap[(\OSS,\theta_{\OSS},\fg\al)]\bigl(\{x\}\bigr)\

3880:       s_{x_{[0,n[}y\inv_{[0,m[}}^{\OSS} s_{\mu\nu\inv}^{\OSS}\

3881:       \boomap[(\OSS,\theta_{\OSS},\fg\al)]

3882:       \bigl(\{\theta_{\nu\mu\inv}^{\OSS}(y)\}\bigr)

3883:       \\

3884:       &= \begin{multlined}[t]

3885:         \boomap[(\OSS,\theta_{\OSS},\fg\al)]\bigl(\{x\}\bigr)\

3886:         \boomap[(\OSS,\theta_{\OSS},\fg\al)]\

3887:         \bigl(D_{x_{[0,n+|\mu|[}y\inv_{[0,m+|\mu|[}}^{\OSS}\bigr)

3888:         \\

3889:         s_{x_{[0,n+|\mu|[}\left(\theta_{vu\inv}(y)_{[0,m+|\nu|[}\right)\inv}^{\OSS}\

3890:         \boomap[(\OSS,\theta_{\OSS},\fg\al)]

3891:         \bigl(\{\theta_{\nu\mu\inv}^{\OSS}(y)\}\bigr)

3892:       \end{multlined}\\

3893:       &= \begin{multlined}[t]

3894:         \boomap[(\OSS,\theta_{\OSS},\fg\al)]\bigl(\{x\}\bigr)\

3895:         s_{x_{[0,n+|\mu|[}(\theta_{vu\inv}(y)_{[0,m+|\nu|[})\inv}^{\OSS}\\

3896:         \boomap[(\OSS,\theta_{\OSS},\fg\al)]

3897:         \bigl(\{\theta_{\nu\mu\inv}^{\OSS}(y)\}\bigr)

3898:       \end{multlined}\\

3899:       &= e^\jj_{x,\theta_{\nu\mu\inv}^{\OSS}(y)}.

3900:     \end{split}

3901:   \end{equation*}

3902:   So we have in all cases that $e^\jj_{x,y} s_g^{\OSS}\in

3903:   \spc\bigl\{e^\jj_{x',y'}\bigm|\jj\in\rte,\

3904:   x',y'\in\spro(\jj)\bigr\}$.

3905: 

3906:   One can in a similar way prove that we have that

3907:   \begin{equation*}

3908:     s_g^{\OSS} e^\jj_{x,y}\in

3909:     \spc\bigl\{e^\jj_{x',y'}\mid\jj\in\rte,\

3910:     x',y'\in\spro(\jj)\bigr\}. 

3911:   \end{equation*}

3912: 

3913:   Hence $\spc\bigl\{e^\jj_{x,y}\mid\jj\in\rte,\

3914:   x,y\in\spro(\jj)\bigr\}$ is an ideal of

3915:   $\csp{\OSS,\theta_{\OSS},\fg\al}$ and thus equal to the kernel of

3916:   $\phi$.

3917: \end{proof}

3918: 

3919: \begin{example} \label{exam:sub}

3920:   Let $\TSS_\eta$ be the shift space of an aperiodic proper

3921:   substitution $\eta$. It then follows from \cite{MR924156}*{page 90

3922:     and 107} and \cite{MR1355295}*{Theorem 3.9} that $\TSS_\eta$ is

3923:   minimal (and thus also $\TSS^+_\eta$) and contains a finite, but

3924:   nonzero, number of left special elements. Thus according to

3925:   \cite{tmcseii}*{Example 3.6}, $\TSS_\eta$ 

3926:   has property $(**)$, and since ${\nTSS}_\eta$ is nonzero, it follows

3927:   from Theorem \ref{theorem:toke} that $\O_{\TSS^+_\eta}$ is not

3928:   simple. On the other hand $C(\TSS_\eta)\rtimes_{\tsh^\star}\Z$ is

3929:   simple because $\TSS_\eta$ is minimal.

3930: \end{example}

3931: 

3932: \appendix

3933: \section{Partial representations of groups} \label{sec:parrep}

3934: A \emph{partial representation} of a group $G$ was defined in

3935: \cite{MR1469405} to be a map $u$ from $G$ to a $\cs$-algebra $\Aa$

3936: such that the following conditions hold:

3937: \begin{subequations}    

3938:   \begin{align}

3939:     &u(e)=1,\label{eq:prxa}\\

3940:     &u(g\inv)=u(g)^* \text{ for every }g\in G\label{eq:prxb},\\

3941:     &u(h)u(i)u(i\inv)=u(hi)u(i\inv) \text{ for every }h,i\in G.\label{eq:prxc}

3942:   \end{align}

3943: \end{subequations}

3944: In \cite{MR99a:46121} another definition of a partial representation

3945: of a group was given, namely as a map $u$ from $G$ to a $\cs$-algebra

3946: $\Aa$ such that the following conditions hold: 

3947: \begin{subequations}

3948:   \begin{align}

3949:     &\parbox[t]{0.8\textwidth}{$\bigl(u(g)\bigr)_{g\in G}$ is a family of partial

3950:       isometries with commuting\\ range projections,}\label{eq:prqra} \\ 

3951:     &u(e)u(e)^*=1,\label{eq:prqrb}\\

3952:     &u(g)^*u(g)=u(g\inv)u(g\inv)^* \text{ for every } g\in G,\label{eq:prqrc}\\

3953:     &u(h)u(i)u(i)^*u(h)^*=u(h)u(i)u(hi)^* \text{ for every } h,i\in

3954:     G.\label{eq:prqrd} 

3955:   \end{align}

3956: \end{subequations}

3957: It was in \cite{MR99a:46121}*{Lemma 1.8 and Remark 1.9} also noticed

3958: that the conditions 

3959: \eqref{eq:prqra}--\eqref{eq:prqrd} are equivalent to the conditions:

3960: \begin{subequations}

3961:   \begin{align}

3962:     &\parbox[t]{0.8\textwidth}{$\bigl(u(g)\bigr)_{g\in G}$ is a family of partial

3963:       isometries with commuting\\ range projections,}\label{eq:pra} \\ 

3964:     &u(e)=1,\label{eq:prb}\\

3965:     &u(g)^*=u(g\inv) \text{ for every } g\in G,\label{eq:prc}\\

3966:     &u(h)u(i)=u(h)u(h)^*u(hi) \text{ for every } h,i\in

3967:     G.\label{eq:prd} 

3968:   \end{align}

3969: \end{subequations}

3970: As mentioned in \cite{MR2003f:46108}, the conditions

3971: \eqref{eq:prxa}--\eqref{eq:prxc} are equivalent to the conditions

3972: \eqref{eq:prqra}--\eqref{eq:prqrd} and thus to the conditions

3973: \eqref{eq:pra}--\eqref{eq:prd}, but since we have not been able to

3974: find a complete proof of this, we will now give one:

3975: \begin{proposition}

3976:   Let $u$ be a map from a group $G$ to a $\cs$-algebra $\Aa$. Then the

3977:   following are equivalent:

3978:   \begin{enumerate}

3979:   \item $u$ satisfies the conditions \eqref{eq:prxa}--\eqref{eq:prxc},

3980:   \item $u$ satisfies the conditions \eqref{eq:prqra}--\eqref{eq:prqrd},

3981:   \item $u$ satisfies the conditions \eqref{eq:pra}--\eqref{eq:prd}.

3982:   \end{enumerate}

3983: \end{proposition}

3984: 

3985: \begin{proof}

3986:   As mentioned above, the equivalence of $(2)$ and $(3)$ follows from

3987:   \cite{MR99a:46121}*{Lemma 1.8 and Remark 1.9}. 

3988: 

3989:   Assume that  $u$ satisfies the conditions

3990:   \eqref{eq:prxa}--\eqref{eq:prxc}. Then we have that

3991:   \begin{equation*}

3992:     u(g)u(g)^*u(g)=u(g)u(g\inv)u(g)=u(e)u(g)=u(g),

3993:   \end{equation*}

3994:   so $u(g)$ is a partial isometry for all $g\in G$.

3995: 

3996:   If $h,i\in G$, then we have that

3997:   \begin{equation*}

3998:     \begin{split}

3999:       u(h)u(h)^*u(i)u(i)^*&=u(h)u(h\inv i)u(i)^*\\

4000:       &=u(i)u(i\inv h)u(h\inv i)u(i)^*\\

4001:       &=\bigl(u(h\inv i)u(i\inv)\bigr)^*\bigl(u(h\inv i)u(i\inv)\bigr)\\

4002:       &=u(i)u(i)^*u(h)u(h)^*u(i)u(i)^*,

4003:     \end{split}

4004:   \end{equation*}

4005:   from which it follows that

4006:   \begin{equation*}

4007:     \begin{split}

4008:       u(i)u(i)^*u(h)u(h)^*&=\bigl(u(h)u(h)^*u(i)u(i)^*\bigr)^*\\

4009:       &=\bigl(u(i)u(i)^*u(h)u(h)^*u(i)u(i)^*\bigr)^*\\

4010:       &=u(i)u(i)^*u(h)u(h)^*u(i)u(i)^*\\

4011:       &=u(h)u(h)^*u(i)u(i)^*.

4012:     \end{split}

4013:   \end{equation*}

4014:   This shows that $u$ satisfies condition

4015:   \eqref{eq:prqra}.

4016: 

4017:   It is clear that $u$ satisfies condition

4018:   \eqref{eq:prqrb} and \eqref{eq:prqrc}. If $h,i\in G$, then we have that

4019:   \begin{equation*}

4020:     \begin{split}

4021:       u(h)u(i)u(i)^*u(h)^*&=u(h)\bigl(u(h)u(i)u(i)^*\bigr)^*\\

4022:       &=u(h)\bigl(u(hi)u(i)^*\bigr)^*\\

4023:       &=u(h)u(i)u(hi)^*,

4024:     \end{split}

4025:   \end{equation*}

4026:   so $u$ also satisfies condition

4027:   \eqref{eq:prqrd}. Thus $(1)$ implies $(2)$.

4028: 

4029:   Assume now that $u$ satisfies the conditions

4030:   \eqref{eq:pra}--\eqref{eq:prd}, and that $h,i\in G$. Then we have that

4031:   \begin{equation*}

4032:     \begin{split}

4033:       u(h)u(i)u(i)^*&=\bigl(u(i)u(i)^*u(h\inv)\bigr)^*\\

4034:       &=\bigl(u(i)u(i\inv h\inv)\bigr)^*\\

4035:       &=u(hi)u(i\inv),

4036:     \end{split}

4037:   \end{equation*}

4038:   which shows that $(3)$ implies $(1)$.

4039: \end{proof}

4040:  

4041: \section{Boolean Algebras} \label{sec:boo}

4042: We recommend \cite{MR29:4713} for a very nice introduction to Boolean

4043: algebras. A \emph{Boolean algebra} is a set $\Bx$ with two

4044: distinct elements $0,1\in\Bx$ which act like the empty set and the

4045: whole set, respectively, and three

4046: operations $\lor:\Bx\times\Bx\to\Bx$, $\land:\Bx\times\Bx\to\Bx$ and

4047: $\neg:\Bx\to\Bx$ which act like union, intersection and complement,

4048: respectively. To be precise, they satisfy the following axioms:

4049: \setlength\arraycolsep{2pt}

4050: \begin{eqnarray}

4051:   \neg 0=1&& \neg 1=0\\

4052:   A\land 0=0&& A\lor 1=1\\

4053:   A\land 1=A&& A\lor 0=0\label{eq:1}\\

4054:   A\land\neg A=0&& A\lor\neg A=1\label{eq:2}\\

4055:   \neg\neg A&=&A\\

4056:   A\land A=A&&A\lor A=A\\

4057:   \neg(A\land B)=\neg A\lor\neg B&& \neg(A\lor B)=\neg A\land\neg B\\

4058:   A\land B=B\land A&& A\lor B=B\lor A\label{eq:3}\\

4059:   A\land(B\land C)=(A\land B)\land C&& A\lor(B\lor C)=(A\lor B)\lor C\\

4060:   A\land(B\lor C)&=&(A\land B)\lor(A\land C)\label{eq:4}\\A\lor(B\land

4061:   C)&=&(A\lor B)\land(A\lor C)\label{eq:5} 

4062: \end{eqnarray}

4063: This set of axioms is not the shortest one possible. In fact one could

4064: do with for example just axiom (\ref{eq:1}), (\ref{eq:2}),

4065: (\ref{eq:3}), (\ref{eq:4}) and (\ref{eq:5}). We will call

4066: $A\lor B$ for the \emph{intersection of $A$ and $B$}, $A\land B$ for the

4067: \emph{union of $A$ and $B$}, and $\neg A$ the \emph{complement of $A$}.

4068: 

4069: The generic example of a Boolean algebra is of course the power set of

4070: a set $X$, where $0=\emptyset$, $1=X$, $\lor=\cup$, $\land=\cap$ and

4071: $\neg A=X\setminus A$. 

4072: 

4073: If $\Aa$ is a subset of a Boolean algebra $\Bx$, then we call it a

4074: \emph{Boolean subalgebra} of $\Bx$ if $0,1\in\Aa$ and $A\lor B, A\land B,

4075: \neg A\in \Aa$ for every $A,B\in \Aa$. In this case $\Aa$ is of course

4076: itself a Boolean algebra with operations inherited from $\Bx$. When

4077: $X$ is a set, then we will by a \emph{Boolean algebra on $X$} mean a Boolean

4078: subalgebra of the power set of $X$. 

4079: 

4080: An example of this which we will use in this paper, is if $X$ is a

4081: topological space. Then the set of clopen subsets of $X$ is a Boolean

4082: subalgebra of the power set of $X$ and thus a Boolean algebra on $X$. 

4083: 

4084: If $\Aa$ is some subset of a Boolean

4085: algebra $\Bx$, then we will by the \emph{Boolean algebra generated by

4086:   $\Aa$} mean the

4087: smallest Boolean subalgebra of $\Bx$ containing $\Aa$. Notice that if a

4088: subset $\Aa$ of a Boolean algebra $\Bx$ is closed under intersection

4089: (respectively union) and complement, then it also closed under union

4090: (respectively intersection) and thus is a Boolean subalgebra.

4091: 

4092: Other examples of Boolean algebras

4093: which we will meet in this paper is the set $\{0,1\}$ where 

4094: \begin{eqnarray*}

4095:   0\lor 0=0\land0=0\land 1=1\land 0=\neg 1=0,\\

4096:   1\land 1=1\lor 1=1\lor 0=0\lor 1=\neg 0=1,

4097: \end{eqnarray*}

4098: and the set of projections in a unital abelian $\cs$-algebra, where

4099: \begin{align*}

4100:   p\lor q=pq&& p\land q=p+q-pq&& \neg p=1-p.

4101: \end{align*}

4102: 

4103: A map $\phi$ between two Boolean algebras $\Bx$ and $\Bx'$ is called 

4104: \emph{a Boolean homomorphism} if $\phi(A\lor B)=\phi(A)\lor\phi(B)$,

4105: $\phi(A\land B)=\phi(A)\land\phi(B)$ and $\phi(\neg A)=\neg\phi(A)$ for

4106: every $A,B\in \Bx$. In fact, the first (respectively the second)

4107: together with the last equality 

4108: imply the second (respectively the first), so in order to verify

4109: that $\phi$ is a Boolean homomorphism, it is enough to check these two

4110: equalities. Notice that when $\phi$ is a Boolean homomorphism, then

4111: $\phi(1)=1$ and $\phi(0)=0$.

4112: 

4113: If $\Aa$ is a unital abelian $\cs$-algebra and $\Bx$ is the Boolean algebra of

4114: projections of $\Aa$, then it is easy to check that $\spa(\Bx)$ is a

4115: $*$-subalgebra of $\Aa$ and thus that $\spc(\Bx)$ is a

4116: $\cs$-subalgebra of $\Aa$. Hence $\spc{\Bx}=\csp{\Bx}$.

4117: 

4118: \begin{lemma} \label{lemma:homo}

4119:   If for $i\in\{1,2\}$, $\Aa_i$ is a unital abelian $\cs$-algebra 

4120:   and $\Bx_i$ is the Boolean algebra of projections of $\Aa_i$, and

4121:   $\phi:\Bx_1\to\Bx_2$ is a Boolean homomorphism, then there is a

4122:   uniquely determined $*$-homomorphism from $\spc{\Bx_1}$ to

4123:   $\spc{\Bx_2}$ which maps $A$ to $\phi(A)$ for $A\in\Bx_1$.

4124: \end{lemma}

4125: 

4126: \begin{proof}

4127:   Since $\Bx_1$ generates $\spc{\Bx_1}$, there can at most be one

4128:   $*$-homomorphism from $\spc{\Bx_1}$ to 

4129:   $\spc{\Bx_2}$ which maps $A$ to $\phi(A)$ for $A\in\Bx_1$.

4130: 

4131:   Let $\cset$ denote the set of finite subset of $\Bx_1$. We then have

4132:   that $\spc{\Bx_1}$ is the closure of the set

4133:   \begin{equation*}

4134:     \bigcup_{C\in\cset}\spa{C},

4135:   \end{equation*}

4136:   and since $\spa{C}$ is the $\cs$-subalgebra of $\Aa_1$ generated by

4137:   $C$, it is enough to show that there for every $C\in \cset$ exists a

4138:   $*$-homomorphism from $\spa{C}$ to

4139:   $\spc{\Bx_2}$ which maps $A$ to $\phi(A)$ for $A\in C$.

4140: 

4141:   So let $C\in\cset$. There then exists a finite family

4142:   $p_1,p_2,\dotsc, p_n$ of mutually orthogonal projections in $C$ such

4143:   that every element of $\spa{C}$ uniquely can be written as

4144:   \begin{equation*}

4145:     \sum_{i_1}^n\lambda_ip_i

4146:   \end{equation*}

4147:   with $\lambda_1,\lambda_2,\dotsc,\lambda_n\in \C$. The map 

4148:   \begin{equation*}

4149:     \sum_{i_1}^n\lambda_ip_i\mapsto \sum_{i_1}^n\lambda_i\phi(p_i)

4150:   \end{equation*}

4151:   is therefore a well-defined $*$-homomorphism from $\spa{C}$ to

4152:   $\spc{\Bx_2}$ which maps $A$ to $\phi(A)$ for $A\in C$.

4153: \end{proof}

4154: 

4155: \section{Crossed products of $\cs$-partial dynamical systems}

4156: \label{sec:parcro} 

4157: 

4158: A \emph{$\cs$-partial dynamical system} has been defined in 

4159: \cite{MR99a:46121} to be a triple $(A,G,\alpha)$ where $A$ is a

4160: $\cs$-algebra, $G$ is a discrete group and $\alpha$ is a 

4161: \emph{partial action} of $G$ on $A$. That means that $\alpha$ consists

4162: of a family $(D_g)_{g\in G}$ of closed ideals of $A$ and a family

4163: $(\alpha_g)_{g\in G}$ of isomorphisms $\alpha_g:D_{g\inv}\to D_g$ such that 

4164: \begin{align}

4165:   &D_e=A,\\

4166:   &\parbox[t]{0.8\textwidth}{$\alpha_{hi}$ extends $\alpha_h\alpha_i$

4167:     for all $h,i\in G$ (where the domain of $\alpha_h\alpha_i$ is

4168:     $\alpha_i\inv(D_{h\inv})$).}

4169: \end{align}

4170: $\cs$-partial dynamical systems have been studied in

4171: \cites{MR2003f:46108,MR99a:46121,MR1331978} (the definition of $\cs$-partial

4172: dynamical systems in \cite{MR1331978} is a bit different from the

4173: above mentioned, 

4174: but it is showed in \cite{MR99a:46121}*{Remark 1.9} that the two

4175: definitions are equivalent). 

4176: 

4177: A \emph{covariant representation} of a $\cs$-partial dynamical system

4178: $(A,G,\alpha)$ on a Hilbert space $\H$ is a pair $(\pi,u)$ where $\pi$

4179: is a non-degenerate representation of $A$ on $\H$, and $u$ is a partial

4180: representation (cf. Appendix \ref{sec:parrep}) of $G$ on $\H$ such

4181: that for each $g\in G$, $u(g)u(g)^*$ is the projection of $\H$ onto

4182: the subspace $\spc\pi(D_g)\H$, and

4183: $\pi\bigl(\alpha_g(a)\bigr)=u(g)\pi(a)u(g\inv)$ for $a\in D_g$. As it

4184: is the case 

4185: with $\cs$-partial dynamical systems, the definition of a covariant

4186: representation in \cite{MR1331978} is a bit different from the above

4187: mentioned, but 

4188: it is shown in \cite{MR99a:46121}*{Remark 1.12} that the two

4189: definitions are equivalent.

4190: 

4191: The \emph{crossed product} $A\rtimes_\alpha G$ of a $\cs$-partial

4192: dynamical system 

4193: $(A,G,\alpha)$ is a $\cs$-algebra which is generated by a copy of $A$

4194: and a family $(\delta_g)_{g\in G}$ of elements such that there exists

4195: a bijective map $(\pi,u)\mapsto \pi\times u$ 

4196: between covariant representations of $(A,G,\alpha)$ on $\H$ and

4197: non-degenerated representations of $A\rtimes_\alpha G$ on $\H$ such

4198: that $(\pi\times u)(a\delta_g)=\pi(a)u(g)$ for $g\in G$ and $a\in D_g$. 

4199: 

4200: \begin{bibdiv} 

4201:   \begin{biblist}

4202:     \bib{MR1355295}{article}{ author={Boyle, Mike}, author={Lind,

4203:         Douglas}, title={Expansive subdynamics},

4204:       journal={Trans. Amer. Math. Soc.}, volume={349}, date={1997},

4205:       number={1}, pages={55\ndash 102}, issn={0002-9947},

4206:       review={MR1355295 (97d:58115)}, }

4207:     

4208:     \bib{phd}{article}{ author={Carlsen, Toke Meier}, title={Operator

4209:         Algebraic Applications in Symbolic Dynamics}, journal={Ph.D

4210:         thesis}, organization={University of Copenhagen}, date={2004},

4211:       pages={189 pp.}, eprint={www.math.ku.dk/~toke}, }

4212:     

4213:     \bib{tmc}{report}{ author={Carlsen, Toke Meier},

4214:       title={Cuntz-Pimsner $C^*$-algebras associated with subshifts},

4215:       status={submitted for publication},

4216:       eprint={arXiv:math.OA/0505503}, note={28 pp.},}

4217:     

4218:     \bib{tmcseii}{article}{ author={Carlsen, Toke Meier},

4219:       author={Eilers, S{\o}ren}, title={Matsumoto $K$-groups

4220:         associated to certain shift spaces}, journal={Doc. Math.},

4221:       volume={9}, date={2004}, pages={639\ndash 671 (electronic)},

4222:       issn={1431-0635}, review={MR2117431 (2005h:37021)}, }

4223:     

4224:     \bib{tmcseiv}{article}{ author={Carlsen, Toke Meier},

4225:       author={Eilers, S\o ren}, title={Ordered {$K$}-groups associated

4226:         to substitutional dynamics}, journal={J. Funct. Anal.},

4227:       volume={238}, number={1}, date={2006}, pages={99\ndash 117} }

4228: 

4229:     \bib{MR2091486}{article}{ author={Carlsen, Toke Meier},

4230:       author={Matsumoto, Kengo}, title={Some remarks on the $C\sp

4231:         *$-algebras associated with subshifts},

4232:       journal={Math. Scand.}, volume={95}, date={2004}, number={1},

4233:       pages={145\ndash 160}, issn={0025-5521}, review={MR2091486

4234:         (2005e:46093)}, }

4235:     

4236:     \bib{CS}{report}{ author={Carlsen, Toke Meier},

4237:       author={Silvestrov, Sergei}, title={$C^*$-crossed products and

4238:         shift spaces}, series={Preprints in Mathematical Sciences

4239:         2005:26}, organization={Lund Institute of Technology},

4240:       date={2005}, note={29 pp.}, eprint={arXiv:math.OA/0512488},

4241:       status={submitted for publication}, }

4242: 

4243:     \bib{MR0467330}{article}{ author={Cuntz, Joachim}, title={Simple

4244:         $C\sp*$-algebras generated by isometries},

4245:       journal={Comm. Math. Phys.}, volume={57}, date={1977},

4246:       number={2}, pages={173\ndash 185}, issn={0010-3616},

4247:       review={MR0467330 (57 \#7189)}, }

4248: 

4249:     \bib{MR561974}{article}{ author={Cuntz, Joachim}, author={Krieger,

4250:         Wolfgang}, title={A class of $C\sp{\ast} $-algebras and

4251:         topological Markov chains}, journal={Invent. Math.},

4252:       volume={56}, date={1980}, number={3}, pages={251\ndash 268},

4253:       issn={0020-9910}, review={MR561974 (82f:46073a)}, }

4254: 

4255:     \bib{MR1469405}{article}{ author={Exel, Ruy}, title={Partial

4256:         actions of groups and actions of inverse semigroups},

4257:       journal={Proc. Amer. Math. Soc.}, volume={126}, date={1998},

4258:       number={12}, pages={3481\ndash 3494}, issn={0002-9939},

4259:       review={MR1469405 (99b:46102)}, }

4260:     

4261:     \bib{MR2000i:46064}{article}{ author={Exel, Ruy}, author={Laca,

4262:         Marcelo}, title={Cuntz-Krieger algebras for infinite

4263:         matrices}, journal={J. Reine Angew. Math.}, volume={512},

4264:       date={1999}, pages={119\ndash 172}, issn={0075-4102}, review={MR

4265:         2000i:46064}, }

4266:       

4267:     \bib{MR2003f:46108}{article}{ author={Exel, Ruy}, author={Laca,

4268:         Marcelo}, author={Quigg, John}, title={Partial dynamical

4269:         systems and $C\sp *$-algebras generated by partial

4270:         isometries}, journal={J. Operator Theory}, volume={47},

4271:       date={2002}, number={1}, pages={169\ndash 186},

4272:       issn={0379-4024}, review={MR 2003f:46108}, }

4273:     

4274:     \bib{MR1363826}{article}{ author={Giordano, Thierry},

4275:       author={Putnam, Ian F.}, author={Skau, Christian F.},

4276:       title={Topological orbit equivalence and $C\sp *$-crossed

4277:         products}, journal={J. Reine Angew. Math.}, volume={469},

4278:       date={1995}, pages={51\ndash 111}, issn={0075-4102},

4279:       review={MR1363826 (97g:46085)}, }

4280: 

4281:     \bib{MR29:4713}{book}{ author={Halmos, Paul R.}, title={Lectures

4282:         on Boolean algebras}, series={Van Nostrand Mathematical

4283:         Studies, No. 1}, publisher={D. Van Nostrand Co., Inc.,

4284:         Princeton, N.J.}, date={1963}, pages={v+147}, review={MR 29

4285:         \#4713}, }

4286: 

4287:     \bib{MR1452183}{article}{ author={an Huef, Astrid},

4288:       author={Raeburn, Iain}, title={The ideal structure of

4289:         Cuntz-Krieger algebras}, journal={Ergodic Theory

4290:         Dynam. Systems}, volume={17}, date={1997}, number={3},

4291:       pages={611\ndash 624}, issn={0143-3857}, review={MR1452183

4292:         (98k:46098)}, }

4293: 

4294:     \bib{MR0346134}{article}{ author={Ito, Shunji}, author={Takahashi,

4295:         Y{\=o}ichir{\=o}}, title={Markov subshifts and realization of

4296:         $\beta $-expansions}, journal={J. Math. Soc. Japan},

4297:       volume={26}, date={1974}, pages={33\ndash 55}, review={MR0346134

4298:         (49 \#10860)}, }

4299: 

4300:     \bib{MR1484730}{book}{ author={Kitchens, Bruce P.},

4301:       title={Symbolic dynamics}, series={Universitext},

4302:       note={One-sided, two-sided and countable state Markov shifts},

4303:       publisher={Springer-Verlag}, place={Berlin}, date={1998},

4304:       pages={x+252}, isbn={3-540-62738-3}, review={MR1484730

4305:         (98k:58079)}, }

4306: 

4307:     \bib{MR561973}{article}{ author={Krieger, Wolfgang}, title={On

4308:         dimension functions and topological Markov chains},

4309:       journal={Invent. Math.}, volume={56}, date={1980}, number={3},

4310:       pages={239\ndash 250}, issn={0020-9910}, review={MR561973

4311:         (81m:28018)}, }

4312:  

4313:     \bib{MR97a:58050}{book}{ author={Lind, Douglas}, author={Marcus,

4314:         Brian}, title={An introduction to symbolic dynamics and

4315:         coding}, publisher={Cambridge University Press},

4316:       place={Cambridge}, date={1995}, pages={xvi+495},

4317:       isbn={0-521-55124-2}, isbn={0-521-55900-6}, review={MR

4318:         97a:58050}, }

4319: 

4320:     \bib{MR1454478}{article}{ author={Matsumoto, Kengo}, title={On

4321:         $C\sp *$-algebras associated with subshifts},

4322:       journal={Internat. J. Math.}, volume={8}, date={1997},

4323:       number={3}, pages={357\ndash 374}, issn={0129-167X},

4324:       review={MR1454478 (98h:46077)}, }

4325:  

4326:     \bib{MR1691469}{article}{ author={Matsumoto, Kengo},

4327:       title={Dimension groups for subshifts and simplicity of the

4328:         associated $C\sp *$-algebras}, journal={J. Math. Soc. Japan},

4329:       volume={51}, date={1999}, number={3}, pages={679\ndash 698},

4330:       issn={0025-5645}, review={MR1691469 (2000d:46082)}, }

4331: 

4332:     \bib{MR1331978}{article}{ author={McClanahan, Kevin},

4333:       title={$K$-theory for partial crossed products by discrete

4334:         groups}, journal={J. Funct. Anal.}, volume={130}, date={1995},

4335:       number={1}, pages={77\ndash 117}, issn={0022-1236},

4336:       review={MR1331978 (96i:46083)}, }

4337: 

4338:     \bib{MR81e:46037}{book}{ author={Pedersen, Gert K.},

4339:       title={$C\sp{\ast} $-algebras and their automorphism groups},

4340:       series={London Mathematical Society Monographs}, volume={14},

4341:       publisher={Academic Press Inc. [Harcourt Brace Jovanovich

4342:         Publishers]}, place={London}, date={1979}, pages={ix+416},

4343:       isbn={0-12-549450-5}, review={MR 81e:46037}, }

4344:     

4345:     \bib{MR924156}{book}{ author={Queff{\'e}lec, Martine},

4346:       title={Substitution dynamical systems---spectral analysis},

4347:       series={Lecture Notes in Mathematics}, volume={1294},

4348:       publisher={Springer-Verlag}, place={Berlin}, date={1987},

4349:       pages={xiv+240}, isbn={3-540-18692-1}, review={MR924156

4350:         (89g:54094)}, }

4351: 

4352:     \bib{MR99a:46121}{article}{ author={Quigg, John}, author={Raeburn,

4353:         Iain}, title={Characterisations of crossed products by partial

4354:         actions}, journal={J. Operator Theory}, volume={37},

4355:       date={1997}, number={2}, pages={311\ndash 340},

4356:       issn={0379-4024}, review={MR 99a:46121}, }

4357: 

4358:     \bib{MR0241994}{article}{ author={Zeller-Meier, G.},

4359:       title={Produits crois\'es d'une $C\sp{\ast} $-alg\`ebre par un

4360:         groupe d'automorphismes}, language={French},

4361:       journal={J. Math. Pures Appl. (9)}, volume={47}, date={1968},

4362:       pages={101\ndash 239}, issn={0021-7824}, review={MR0241994 (39

4363:         \#3329)}, }

4364: 

4365:   \end{biblist}

4366: \end{bibdiv}

4367: \end{document}

4368: 

4369: 

4370: %%% Local Variables: 

4371: %%% mode: latex

4372: %%% TeX-master: t

4373: %%% End: 

4374: