math0306354/cod.tex
1: \documentclass{amsart}
2: \usepackage{amscd,amssymb}
3: \usepackage[dvips]{graphicx}
4: 
5: \numberwithin{equation}{section}
6: 
7: \newtheorem{thm}{Theorem}
8: \newtheorem{co}[thm]{Corollary}%[section]
9: \newtheorem{lem}[thm]{Lemma}
10: \newtheorem{pro}[thm]{Proposition}
11: \theoremstyle{remark}
12: 
13: \newtheorem{rem}[thm]{Remark}
14: \newtheorem{defi}[thm]{Definition}
15: \newtheorem{exa}[thm]{Example}
16: \newtheorem{nota}[thm]{Notation}
17: 
18: 
19: 
20: \begin{document}
21: \title{Coding and tiling of Julia sets 
22: for subhyperbolic rational maps }
23: \author{Atsushi KAMEYAMA}
24: \address{Division of Mathematical Science for Social Systems\\
25: Department of Systems Innovation\\
26: Graduate School of Engineering Science\\
27: Osaka University, Toyonaka, Osaka, 560-8531, Japan}
28: \email{kameyama@sigmath.es.osaka-u.ac.jp}
29: \urladdr{http://www.sigmath.es.osaka-u.ac.jp/\~{}kameyama/}
30: \date{}
31: 
32: \maketitle
33: 
34: \begin{abstract}
35:  Let $f:\hat{\mathbb C}\to\hat{\mathbb C}$ be a subhyperbolic  
36: rational map of degree $d$.
37: We construct a set of coding maps $\mathrm{Cod}(f)=
38: \{\pi_r:\Sigma\to J\}_r$ of 
39: the Julia set $J$ by 
40: geometric coding trees, where the parameter $r$ ranges over 
41: mappings from a certain tree to the Riemann sphere.
42: Using the universal covering space $\phi:\tilde S\to S$ 
43: for the corresponding orbifold, 
44: we lift the inverse of $f$  to an iterated function system 
45: $\mathcal I=(g_i)_{i=1,2,\dots,d}$. 
46: For the purpose of studying the structure of $\mathrm{Cod}(f)$, 
47: we generalize Kenyon and Lagarias-Wang's results 
48: : If the attractor $K$ of $\mathcal I$ 
49: has positive measure, then $K$ tiles $\phi^{-1}(J)$, and the 
50: multiplicity of $\pi_r$ is well-defined.
51: Moreover, we see that the equivalence relation induced by $\pi_r$ 
52: is described  by a finite directed graph, and give a necessary and 
53: sufficient condition for two coding maps $\pi_r$ and $\pi_{r'}$ to 
54: be equal.
55: \end{abstract}
56: 
57: \section{Introduction}
58: 
59: The method of symbolic dynamics is prevalent in the study
60:  of dynamical systems.
61: Particularly, to investigate an attractor (or repeller) of 
62: a dynamical system,
63: one often uses symbolic dynamics to code the attractor.
64: In the present paper, we study coding maps of Julia sets for rational 
65: maps.
66: We emphasize that we treat not only individual coding maps but 
67: totalities of coding maps.
68: 
69: 
70: We say a rational map $f:\hat{\mathbb{C}}\to\hat{\mathbb{C}}$ 
71: of the Riemann sphere to itself is subhyperbolic if each critical point 
72: is either preperiodic or attracted to an attracting cycle.
73: We will construct a set of coding maps $\mathrm{Cod}(f)=\{\pi_r\}_r$ 
74: from the full 
75: shift to the Julia set $J$ by using `geometric coding trees.' 
76: Geometric coding tree technique is developed by Przytycki and 
77: his coauthors for 
78: general holomorphic maps (\cite{Prz85}, \cite{Prz86}, \cite{PrSk91}).
79: See also \cite{Ly86}, \S 1.16.
80: The parameter $r$ ranges over  {\em radials}, 
81: which are mappings from 
82: a certain topological tree to the Riemann sphere. 
83: One goal of considering $\mathrm{Cod}(f)$ is to understand the
84: combinatorics of $f$.
85: In view of the construction of geometric coding trees, the structure 
86: of $\mathrm{Cod}(f)$ reflects the combinatorics of $f$.
87: 
88: We know that coding is effectively used in the study of 
89: dynamics of interval maps (see for example \cite{CoEc80}, \cite{MiTh88}, 
90: \cite{MeSt93}).
91: Coding often works in the parameter space as well as in the
92: dynamical space.
93: Recall that the dynamical space is just an interval $I$ with natural 
94: partition, that is, $I$ is divided 
95: into several subintervals by the turning points.
96: This partition gives a natural coding of $I$, and we obtain a nice 
97: invariant called the kneading sequence, which is 
98: defined as the symbol sequences corresponding to the forward orbits 
99: of the turning points.
100: For example, in a certain family of real polynomial maps the 
101: kneading sequences almost completely classifies these maps 
102: up to topological conjugacy. 
103: In particular, for the quadratic family $x\mapsto ax(x-1)$ 
104: ($0<a<4$), we have the monotonicity of the kneading sequence (and 
105: the topological entropy).   
106: Roughly speaking, the natural coding parametrizes the bifurcation of the 
107: quadratic family. 
108: However, in a larger family of complex rational maps, 
109: coding does not seem to work well in the parameter space.
110: The main reason of this difficulty is the absence of natural partitions.
111: We do not have a nice invariant like the kneading sequence.
112: In such a situation, thus it is less important to consider individual 
113: coding maps.
114: This is why we treat totalities of coding maps.
115: 
116: A complete description of $\mathrm{Cod}(f)$ is quite difficult except 
117: a few cases including $f(z)=z^2$, $f(z)=z^2-2$, etc.
118: It is unfortunate that even the case $f(z)=z^d$ with $d\ge 3$ 
119: and the Cantor set case (e.g. $f(z)=z^2-3$) are complicated.
120: Thus we will try to find  tools to manage
121: $\mathrm{Cod}(f)$, keeping in mind the following natural and 
122: naive problems: (1) What is the canonical coding?
123: (2) Are there any good structures on $\mathrm{Cod}(f)$?
124: The present paper does not completely solve these problem, 
125: but gives several fundamental facts which will be a help to approach 
126: these problems.
127: Our main results are concerned with the 
128: multiplicities of coding maps and the equivalence relations on the 
129: space of symbol sequences.
130: 
131:  
132: 
133: 
134: In our setting, $f^{-1}$ can be lifted by the `universal 
135: covering' $\phi:\tilde S\to S=\hat{\mathbb C}-\mathrm{AP}$, 
136: where $\mathrm{AP}$ is the set of attracting periodic points.
137: Let $d$ be the degree of $f$.
138: For $i=1,2,\dots,d$, there exists a holomorphic contraction
139: $g_i:\tilde S\to \tilde S$ depending on $r$ such that the diagram
140: 
141: $$\begin{CD}
142:    \tilde S-\phi^{-1}f^{-1}(\mathrm{AP}) @<{g_i}<< \tilde S\\
143:    @V\phi VV         @VV\phi V \\
144:    S-f^{-1}(\mathrm{AP})       @>> f >       S
145:   \end{CD}
146: $$
147: commutes.
148: Thus there exists a compact set $K$ such that $K=\bigcup_{i=1}^d 
149: g_i(K)$ by Hutchinson \cite{Hu81}.
150: We call $K$ the {\em Julia tile} with respect to $r$.
151: If $r$ is `suitable', then the coding map $\pi_r$ is onto, and so 
152: $\phi(K)=J$. 
153: 
154: The situation above is analogous to that of self-affine tiling.
155: We recall self-affine tiling briefly.
156: Let $A$ be an $n\times n$ expanding integral matrix, and let 
157: $d=|\det A|$.
158: Since $A\mathbb{Z}^n\subset\mathbb{Z}^n$, its projection $f:\mathbb T^n
159: \to\mathbb T^n$ is well-defined, where $\mathbb T^n=\mathbb R^n/\mathbb 
160: Z^n$ is the $n$-torus.
161: A lift of $f$ has the form $x\mapsto Ax+a,a\in \mathbb Z^n$.
162: Choosing $a_1,a_2,\dots,a_d\in\mathbb Z^n$, we obtain $d$ contractions
163: $g_i:x\mapsto A^{-1}(x-a_i)$ by which we have the commuting diagram 
164: $$\begin{CD}
165:    \mathbb R^n @<{g_i}<< \mathbb R^n\\
166:    @V\phi VV         @VV\phi V \\
167:    \mathbb T^n       @>> A >       \mathbb T^n
168:   \end{CD}
169: $$
170: for $i=1,2,\dots,d$, and the self-affine set $K$ with $K=\bigcup_{i=1}^d 
171: g_i(K)$.
172: Kenyon \cite{Ken92} showed that if $K$ has positive Lebesgue measure, 
173: then $K$ tiles $\mathbb R^n$ (i.e. there exists a set of translations 
174: $T\subset \mathbb Z^n$ such that (1) $\mathbb R^n=\bigcup_{t\in T}(K+t)$, 
175: and (2) the Lebesgue measure of $(K+t)\cap (K+t')$ vanishes for distinct 
176:  $t,t'\in T$).
177: See for more details \cite{Ken92} and \cite{LaWa96b}.
178: 
179: We generalize Kenyon and Lagarias-Wang's results and obtain:
180: 
181: \medskip
182: 
183: \noindent{\bf Tiling Theorem} (Theorem \ref{thm:2}).
184: Let $\mu$ be the equilibrium state for $f:J\to J$ with constant 
185: weight, and $\tilde\mu$ the lift of $\mu$ to $\tilde S$.
186: If $\tilde\mu(K)$ is positive, then $K$ tiles $\tilde J=\phi^{-1}(J)$, 
187: that is, there exists a set $T$ of deck 
188: transformations of $\phi:\tilde S\to S$ such that  
189:  (1) $\tilde J=\bigcup_{t\in T}t(K)$, 
190: and (2) $\tilde\mu(t(K)\cap t'(K))=0$ for distinct  $t,t'\in T$.
191: 
192: \medskip
193: 
194: \noindent{\bf Multiplicity Theorem} (Theorem \ref{thm:21}).
195: If $\tilde\mu(K)$ is positive, then there exists $n>0$ depending on 
196: $r$ such that $\pi_r$ is {\em almost $n$-to-one}, 
197: that is, $\#\pi_r^{-1}\pi_r(\omega)=n$ for $m$-almost 
198: all $\omega\in\Sigma$, where $m$ is the Bernoulli measure with
199: identically distributed weight. 
200: The number $n=n_r$ is said to be the multiplicity of $\pi_r$.
201: \medskip
202: 
203: For a given coding map $\pi_r$, it is difficult to calculate its 
204: multiplicity in general.
205: The most primitive way to do that is directly seeing the equivalence 
206: relation induced by $\pi_r$ on $\Sigma$.
207: We will show that the equivalence relation is described by a finite 
208: graph, which is a version of Fried's result (\cite{Fri87}, Lemma 1).
209: See also \cite{Ka03}.
210: 
211: \medskip 
212: 
213: \noindent{\bf Finite Graph Theorem} (Theorem \ref{thm:28}). 
214: We can construct a directed graph with vertex set $V$, edge set $E$, and 
215: weight $\alpha:E\to \{1,2,\dots,d\}^2$ such that 
216: $\pi_r(\omega_1\omega_2\cdots)=\pi_r(\omega'_1\omega'_2\cdots)$ 
217: if and only if there exists a sequence $e_1,e_2,\ldots\in E$ with 
218: $\alpha(e_i)=(\omega_i,\omega'_i)$ and $(\text{the terminal vertex of 
219: $e_i$})=(\text{the initial vertex of $e_{i+1}$})$ for $i=1,2,\dots$.
220: 
221: \medskip
222: 
223: We expect that any subhyperbolic rational maps have the canonical 
224: coding maps.
225: The word `canonical' has vagueness, but it become clearer by the notion 
226: of multiplicity.
227: If there exists the simplest nontrivial coding map, we may consider it 
228: canonical.
229: We will look at several examples of subhyperbolic rational maps later,
230: and will find out that each of them has a nontrivial 
231: coding map which is apparently the simplest.
232: These coding maps have the features (1) the multiplicities are equal to 
233: one, (2) the {\em connecting sets} 
234: $$E=\bigcup_{i\ne j}(g_i(K)\cap g_j(K))
235: \cup\bigcup_i \{x\in g_i(K)\,|\,\#g_i^{-1}(x)\ge 2\}$$
236: are small, and (3) the Julia tiles $K$ have simple shapes.
237: While the second and the third features are still vague, the first one 
238: mathematically makes sense.
239: Thus our conjecture is that any subhyperbolic rational maps have a  
240: coding maps with multiplicity one.
241: 
242: Another conjecture is that $\mathrm{Cod}(f)$ has some structure.
243: For example, we expect that there exists a natural action on 
244: $\mathrm{Cod}(f)$ by which we can control the diversity of multiplicities 
245: of $\pi_r$.
246: 
247: \medskip
248: \noindent{\bf Structure Theorem} (Theorem \ref{thm:24}).
249: $\pi_r=\pi_{r'}$ if and only if $r$ and $r'$ are freely homotopic mod 
250: $N_r$.
251: 
252: \medskip
253: This theorem says that 
254: $$\mathrm{Cod}(f)\approx
255: \bigcup_N\{r\,|\,N_r=N\}/(\text{freely homotopic mod $N$}),$$ 
256: where $N_r$ denotes the maximal invariant subgroup with respect to $r$ 
257: of the fundamental group of $\mathbb C-\{\text{postcritical points}\}$ 
258: we will define later.
259: Let $\mathcal A(f)$ be the monoid of rational maps commutative to $f$.
260: Then $\mathcal A(f)$ acts on $\mathrm{Cod}(f)$ by 
261: $(R,\pi)\mapsto R\circ\pi$, and the multiplicity of $R\circ\pi$ is 
262: equal to the degree of $R$ times the multiplicity of $\pi$ 
263: (Proposition \ref{pro:32}).
264: 
265: 
266: 
267: The organization of the present paper is as follows.
268: After giving the definitions of coding maps and Julia tiles in Section 
269: \ref{sec:def}, we consider several examples of subhyperbolic rational 
270: maps in Section \ref{sec:3}.
271: In one example, the L\'evy Dragon appears as a Julia tile.
272: Section \ref{sec:4} supplies the definitions of invariant subgroups 
273: and the equilibrium states, and shows some basic facts.
274: We prove Tiling Theorem and Multiplicity Theorem in Section 
275: \ref{sec:tile}.
276: In Section \ref{sec:st}, we discuss the structure of $\mathrm{Cod}(f)$, 
277: prove Structure Theorem, and show a couple of examples. 
278: Finite Graph Theorem  is proved in Section \ref{sec:5}.
279: In Section \ref{sec:nonhyp}, we see that several results hold for 
280: non-subhyperbolic rational maps.
281: 
282: \section{Definition}\label{sec:def}
283: 
284: In this section, we give the construction of coding maps after 
285: Przytycki.  
286: We obtain Julia tiles by lifting the coding maps to the universal 
287: covering spaces.
288: Let $f:\hat{\mathbb C}\to\hat{\mathbb C}$ be a rational map of degree 
289: $d$. 
290: 
291: \begin{defi}
292: We say that $$C=C_f=\{\text{critical points of $f$}\}
293: =\{c\,|\,\text{$f$ is not locally homeomorphic at $c$}
294: \}$$ is the {\em critical set} of $f$. 
295: The {\em postcritical set} is defined by 
296: $$P=P_f=\overline{\{f^k(c)\,|\,c\in C,n>0\}}.$$
297: Let $\mathrm{AP}$ denote the set of attracting periodic points of $f$.
298: We write $$S'=\hat{\mathbb{C}}-P\quad\text{and}\quad
299:  S=\hat{\mathbb{C}}-\mathrm{AP}.$$
300: \end{defi}
301: 
302: 
303: \begin{defi}
304:  Let $$Q=Q_d=\bigcup_{i=1}^d[0,1]_i/(0_i\sim 0_j)$$
305: be the topological tree made of $d$ copies of the unit interval 
306: with all the origins identified.
307: We say that a continuous map $r:Q\to S'$ is a 
308: {\em radial} if $r(0)=f\circ r(1_i)$ for $i=1,2,\dots,d$.
309: We say that $\bar x=r(0)$ is the {\em basepoint} of $r$.
310: A radial is considered as a $d$-tuple of curves $(l_i:[0,1]\to S')_i$ 
311: with the same initial point $\bar x$ such that $f\circ l_i(1)=\bar x$.
312: A radial $r$ is said to be {\em proper} if $r(1_i)\ne r(1_j)$ whenever
313: $i\ne j$.
314: We write the set of radials with basepoint $\bar x$ by 
315: $$\mathrm{Rad}(f,\bar x). $$
316:  Set 
317: $$\begin{array}{l}
318: L=L(f,{\bar x})=\{l:[0,1]\to S'\,|\,\text{$l$ is continuous and 
319: $l(0)=\bar x$}\},\\
320: \Lambda(f,\bar x)=\{l\in L(f,\bar x)\,|\,l(1)\in f^{-1}(\bar x)\}.
321: \end{array}$$
322: Then $\mathrm{Rad}(f,\bar x)=\Lambda(f,\bar x)^d$.
323: 
324: For a curve $l\in L$ and $x\in f^{-k}(\bar x)$, 
325: we define a curve $F_x(l):[0,1]\to S$ as the lift of $l$ by $f^k$ 
326: with initial point $x$, that is, $f^k\circ F_x(l)=l$ and $F_x(l)(0)=x$.  
327: Since $l$ does not pass through $P$, the curve $F_x(l)$ is uniquely 
328: defined.
329: \end{defi}
330: 
331: \begin{nota}
332: Let $(\sigma,\Sigma)$ be the one-sided fullshift of $d$ symbols.
333: Namely, $$\Sigma=\{1,2,\dots,d\}^{\mathbb{N}}$$ is the set of 
334: one-sided infinite sequences of $\{1,2,\dots,d\}$, and $\sigma:\Sigma\to
335: \Sigma$ is the shift map ($\sigma(w_1w_2\cdots)=w_2\cdots$).
336: 
337: Let $W$ be the set of words (i.e. finite sequences) of $d$ symbols, 
338: and $W_k$ the set of words of length $k$:
339: $$W=\bigcup_{k=1}^\infty W_k,\quad W_k=\{1,2,\dots,d\}^k.$$ 
340: For $w\in W$, we write 
341: $$\Sigma(w)=\{w\omega_1\omega_2\cdots\,|\,\omega_1\omega_2
342: \cdots\in\Sigma\}\subset\Sigma. $$
343: \end{nota}
344: 
345: Construction of coding maps.
346: Suppose $f$ is subhyperbolic.
347: For a radial $r=(l_i)_i$, we inductively define curves $l_w:[0,1]\to 
348: S'$ and points $x_w$ for $w\in W$.
349: First we set $x_i=l_i(1)$ for $i\in W_1=\{1,2,\dots,d\}$.
350: If $l_w$ and $x_w$ are determined for $w\in W_k$, 
351: we set $l_{iw}=l_i\cdot F_{x_i}(l_w)$ and 
352: $x_{iw}=l_{iw}(1)$ for $i\in W_1$.
353: By the expandingness of $f$, 
354: we have $l_\omega=\text{``$\lim_{k\to\infty}$''}
355: l_{\omega_1\omega_2\cdots\omega_k}$ and $x_\omega=\lim_{k\to\infty}
356: x_{\omega_1\omega_2\cdots \omega_k}$ for $\omega=\omega_1\omega_2\cdots
357: \in\Sigma$.
358: Note that $l_{\omega_1\omega_2\dots\omega_k}$ does not converge, but 
359: by a suitable change of parametrization, the limit $l_\omega$ is 
360: well-defined and unique up to parametrization.
361: Clearly, $x_\omega=l_\omega(1)$ for $\omega\in\Sigma$.
362: Since $x_\omega$ is an accumulation point of $f^{-k}(\bar
363: x),k=1,2,\dots$, the point belongs to the Julia set (for example, 
364: see \cite{Mi99}, 4.7).
365: It is easily seen that $f(x_\omega)=x_{\sigma\omega}$ and 
366: the mapping $\omega\mapsto x_\omega$ is continuous.
367: We denote this mapping by $\pi=\pi_r:\Sigma\to J$, and call it 
368: the {\em coding map} of $J$ for $r$.
369: We write $$\mathrm{Cod}(f)=\{\pi_r\,|\, \text{$r$ is a radial}\}.$$
370: 
371: \begin{rem}\label{rem:5}
372: \ 
373: \begin{itemize}
374: \item
375: $l_{u}\cdot F_{x_u}(l_w)$ is equal to $l_{uw}$ 
376: up to parametrization for $u\in W$ and $w\in W$ (or $w\in\Sigma$). 
377: \item $f\pi(\Sigma)=\pi(\Sigma)$.
378: \item If $r$ is proper, then $\pi_r:\Sigma\to J$ is onto.
379: However, the converse is not always true.
380: 
381: \item
382:  The image of $\pi$ is either a perfect set or a singleton.
383: Indeed, suppose $\pi(\Sigma)$ has a isolated point $p$.
384: Since $\pi^{-1}(p)$ is open, there exists $k$ such that 
385: $\sigma^k\pi^{-1}(p)=\Sigma$.
386: Therefore $\pi(\Sigma)=f^k(p)$.
387: 
388: \item
389: If $\pi(\Sigma)$ is perfect, then $\#f^{-k}(x)\cap\pi(\Sigma)\to\infty$ 
390: as $k\to\infty$ for any $x\in\pi(\Sigma)$.
391: Indeed, it is sufficient to show that there 
392: exists $k$ such that $\#f^{-k}(x)\cap\pi(\Sigma)\ge 2$.
393: Assume that $f^{-k}(x)\cap\pi(\Sigma)=\{y_k\}$ for any $k$.
394: There exists $k_0$ such that $y_k\notin C$ for $k\ge k_0$.
395: Take $\omega\in\Sigma$ with $\pi(\omega)=y_{k_0}$.
396: Then $\pi(w\omega)=y_{k_0+m}$ for any $w\in W_m$.
397: This means that $F_{x_w}(l_\omega)$'s are the same for $w\in W_m$.
398: Thus $x_w$'s are the same for $w\in W_m$.
399: Consequently, the accumulation points of $\{x_w\}_{w\in W_m}$ ($m\to
400: \infty$) consist of one point.
401: 
402: \end{itemize}
403: \end{rem}
404: 
405: From now on, we suppose that $f$ is subhyperbolic.
406: 
407: \begin{defi}\label{def:3}
408: A function $\rho:S\to\mathbb{N}$ is called a {\em ramification 
409: function} for $f$ 
410: if $\rho(x)=1$ for $x\notin P$ and $\rho(f(x))$ is a multiple 
411: of $\deg_xf\cdot\rho(x)$.
412: The minimal ramification function is called the {\em canonical} 
413:  ramification function and denoted by $\rho_f$. 
414: %(Note that the canonical ramification function is well-defined for 
415: %general rational maps $f$.)
416: 
417: For a ramification function $\rho$, we have a {\em universal covering} 
418: $\phi:\tilde S\to S$ for the {\em orbifold} $(S,\rho)$ (i.e. 
419: $\tilde S$ is a connected and simply connected Riemann surface and 
420: $\phi$ is the holomorphic branched covering such that the local degree 
421: $\deg_{\tilde x} \phi$ is $\rho(\phi(\tilde x))$ for every 
422: $\tilde x\in S$).
423: See \cite{Mi99}, Appendix E.
424: 
425:  The universal covering $\phi:\tilde S\to S$ is constructed as 
426: follows.
427: Let $$G=\pi_1(S',\bar x)$$
428: be the fundamental group of $S'$.
429: For $$\gamma\in\Gamma=\Gamma(f,\bar x)
430: =\{\gamma\in L(f,\bar x)\,|\,\gamma(0)=\gamma(1)\},$$
431: we denote the homotopy class for $\gamma$ by $[\gamma]\in G$.
432: Let $\mathrm{AP}=\{a_1,a_2,\dots,a_p\}$ and $P-\mathrm{AP}=\{b_1,b_2,
433: \dots\}$.
434: Choose simple closed curves 
435: $A_1,A_2,\dots,A_{p},B_1, B_2,\dots\in\Gamma$ so that 
436: $A_i$ separates $a_i$ from the other points of $\mathrm{AP}$, and 
437: $B_j$ separates $b_j$ from the other points of $P$.
438: Then $G$ is generated by $[A_i],[B_j]\ (i=1,2,\dots,p,\ j=1,2,\dots)$.
439: Let $N=N^\rho$ be the normal subgroup of $G$ generated by 
440: $[B_j^{\rho(b_j)}],b_j\in P-\mathrm{AP}$. 
441: Then $[A_i^k]\notin N,k\in\mathbb Z$ and $[B_j^k]\notin N,
442: 0<|k|<\rho(b_j)$ in our case.
443: The quotient group 
444: $$G^\rho=G/N$$
445: is called the fundamental group of the orbifold $(S,\rho)$.
446:  Let $$\bar L
447: =\{l:[0,1]\to S\,|\,\text{$l$ is continuous, $l(0)=\bar x$ 
448:  and $l(t)\in S'$
449: for $0\leq t<1$}\}.$$
450: We set $$\tilde S=\bar L/\sim_N,$$
451: where $l\sim_N l'$ if $l(1)=l'(1)$ and there exists a closed curve 
452: $\gamma\in \Gamma$ obtained by perturbing $ll'^{-1}$ near $l(1)$  
453: such that $[\gamma]\in N$. 
454: We obtain a branched covering $\phi:\tilde S\to S$ defined by 
455: $\phi([l])=l(1)$.
456: The surface $\tilde S$ is considered as either $\{|z|<1\}$ or 
457: $\mathbb{C}$.
458: 
459: The covering $\phi:\tilde S-\phi^{-1}(P)\to S'$ corresponds to 
460: the normal subgroup $N\subset G$.
461: Take $\tilde x\in\phi^{-1}(\bar x)$.
462: Since $(f\circ \phi)_*\pi_1(\tilde S-\phi^{-1}f^{-1}(P),z)
463: \supset N$, 
464: for $z\in \phi^{-1}f^{-1}(\bar x)$ 
465: there exists a holomorphic covering $g:\tilde S-\phi^{-1}(P)\to 
466: \tilde S-\phi^{-1}f^{-1}(P)$ such that 
467: \begin{equation}
468: f\circ \phi\circ g=\phi\label{eq:0}
469: \end{equation}
470:  and $g(\tilde x)=z$.
471: We extend $g$ to a holomorphic branched covering $g:\tilde S\to\tilde S$.
472: We say that $g$ is the {\em contraction associated with} $z$ (with 
473: respect to the basepoint $\tilde x$). 
474: 
475: By retaking a bigger ramification function if necessary, we may 
476: assume $\tilde S=\{|z|<1\}$. 
477: Thus $g$ is contracting with respect to the Poincar\'e metric.
478: Hence $f$ is uniformly expanding near $J$ with respect to the 
479: projection metric, that is, 
480: there exists $c>1$ such that $\|Df_z\|>c$ for $z$ in some neighborhood 
481: of $J$.
482: See \cite{Mi99}, \S19.
483: \end{defi}
484: 
485: 
486: 
487: 
488: \begin{defi}\label{def:8}
489: Fix $\tilde x\in \phi^{-1}(\bar x)$.
490: For $l\in \bar L$, we denote by $\tilde l$ 
491: the lift of $l$ to $\tilde S$ with $\tilde l(0)=\tilde x$
492: Let $r=(l_i)$ be a radial.
493: We have the contractions $g_1,g_2,\dots,g_d:\tilde S\to\tilde S$ 
494: associated with $\tilde l_i(1)$. 
495: Since $g_i$'s are contracting, we have the attractor $K$ of the iterated 
496: function system $\mathcal I=(g_1,g_2,\dots,g_d)$
497: (i.e. $K\subset \tilde S$ is the unique nonempty compact set 
498: with $K=\bigcup_{i=1}^dg_i(K)$. See \cite{Hu81}).
499: We use the notation 
500: $$g_w=g_{w_1}\circ g_{w_2}\circ\cdots\circ g_{w_k}$$
501: for $w=w_1w_2\cdots w_k\in W$.
502: It is known that a surjective coding map $$\tilde\pi:\Sigma\to K$$
503: is defined by $\tilde\pi(\omega_1\omega_2\cdots)=\lim_{k\to\infty}
504: g_{\omega_1\omega_2\cdots\omega_k}(z)$, which is independent of 
505: $z\in \tilde S$. 
506: \end{defi}
507: 
508: \begin{pro}\label{pro:ss}
509: $\phi\circ\tilde\pi=\pi$.
510: \end{pro}
511: 
512: \begin{proof}
513: We show that $\tilde l_w(1)=g_w(\tilde x)$ for any $w\in W$ 
514: by induction.
515: Assume that $\tilde l_w(1)=g_w(\tilde x)$ for $w\in W_k$.  
516: Then $\phi\circ g_w(\tilde x)=x_w$.
517: By (\ref{eq:0}), $f^k\circ \phi\circ g_w\circ \tilde l_i=l_i$.
518: Hence $F_{x_{w}}(l_i)=\phi\circ g_w\circ \tilde l_i$.
519: Therefore $l_{wi}=l_w\cdot(\phi\circ g_w\circ \tilde l_i)$ 
520: lifts to $\tilde l_w\cdot (g_w\circ \tilde l_i)$.
521: Hence $\tilde l_{wi}(1)=g_w\circ \tilde l_i(1)=g_{wi}(\tilde x)$.
522: 
523: From $\tilde \pi(w_1w_2\cdots)=\lim_{k\to\infty}g_{w_1w_2\cdots w_k}
524: (\tilde x)$ and $g_w(\tilde x)=\tilde l_w(1)$, 
525: it follows that $\phi\circ \tilde \pi(w_1w_2\cdots)
526: =\lim_{k\to\infty}x_{w_1w_2\dots w_k}=\pi(w_1w_2\cdots)$.
527: \end{proof}
528: 
529: \begin{rem}
530: Let $$[r]_N=(\tilde l_i(1))\in(\phi^{-1}f^{-1}(\bar x))^d.$$
531:  Since the iterated function system $\mathcal I$ is determined by 
532: $(\tilde l_i(1))_i$, 
533: so is the coding map $\pi_r$. 
534: \end{rem}
535: 
536: 
537: \section{Examples}\label{sec:3}
538: 
539: \subsection{}
540: Let $f(z)=z^d$.
541: Then $C=P=\mathrm{AP}=\{0,\infty\}$, $S=\mathbb C-\{0\}$, and 
542: the Julia set $J$ is the unit circle $\{|z|=1\}$.
543: Since $P-\mathrm{AP}=\emptyset$, we have a unique ramification function 
544: $\rho(x)=1$, and so $N=N^\rho$ is trivial.
545: Thus $\tilde S$ is Euclidean (i.e. $\tilde S=\mathbb C$).
546: We take an universal covering $\phi=z\mapsto e^{-2\pi iz}:\mathbb C\to 
547: \mathbb C-\{0\}$. 
548: Fix basepoints $\bar x=1\in\mathbb C-\{0\}$ and $\tilde x=0\in
549: \mathbb C$.
550: Then $\phi^{-1}f^{-1}(\bar x)=\frac1d\mathbb Z$.
551: The contraction associated with $n/d\in\frac1d\mathbb Z$ is 
552: $z\mapsto z/d+n/d$.
553: The coding maps $\pi_r$ such that $[r]_N=(n_1/d,n_2/d,\dots,n_d/d)$ with 
554: $\{n_1,n_2,\dots,n_d\}=\{0,1,\dots,d-1\}\mod d$, 
555: for which the attractors $K_r$ are intervals of length one, 
556: are considered canonical.
557: 
558: Consider the case $d=2$.
559: If $[r]_N=(n_1/2,n_2/2)$ with $n_1\le n_2$, 
560: then $K_r$ is the closed interval $[n_1,n_2]$, and $\pi_r$ 
561: is almost $(n_2-n_1)$-to-one map provided $n_2-n_1>0$.
562: Since $K_r$ is a closed interval, $K_r$ tiles $\phi^{-1}(J)=\mathbb R$ 
563: whenever $n_2-n_1$ is positive. 
564: 
565: Consider the case $d=3$.
566: Let $r$ be a radial with $[r]_N=(n_1/3,n_2/3,n_3/3)=(km_1/3,km_2/3,km_3/3)
567: $, where $k$ is the greatest common divisor of $n_1,n_2$, and $n_3$. 
568: Kenyon \cite{Ken97} proved that $\mu(K_r)$, 
569: the 1-dimensional Lebesgue measure of 
570: $K_r$ is positive if and only if $m_1+m_2+m_3=0\mod 3$, and that then 
571: $\mu(K_r)=k$ and $K_r$ tiles $\phi^{-1}(J)=\mathbb R$.
572: Remark that $K_r$ is not necessarily an interval.
573: Related topics are discussed by Lagarias and Wang \cite{LaWa96}.
574: 
575: \subsection{}
576: Let $f(z)=z^2-2$.
577: Then $C=\{0,\infty\}$, $P=\{-2,2,\infty\}$, $\mathrm{AP}=\{\infty\}$, 
578: $S=\mathbb C$, and 
579: the Julia set $J$ is the interval $[-2,2]$.
580: The canonical ramification function is $\rho(z)=2$ if $z=-2,2$, 
581: $\rho(z)=1$ otherwise.
582: Note that $\tilde S$ is Euclidean, 
583: and take a universal branched covering $\phi=z\mapsto 2\cos\pi z:
584: \mathbb C\to\mathbb C$.
585: Then $\phi^{-1}(J)$ is the real axis and $f\circ\phi(z)=\phi(\pm 2z+2n),
586: n\in\mathbb Z$.
587: Fix basepoints $\bar x=0$ and $\tilde x=1/2$.
588: Then $f^{-1}(\bar x)=\{\pm\sqrt{2}\}$, 
589: $\phi^{-1}(\sqrt{2})=\{1/4+2n,7/4+2n\,|\,n\in\mathbb Z\}$ 
590: and $\phi^{-1}(-\sqrt{2})=\{3/4+2n,5/4+2n\,|\,n\in\mathbb Z\}$.
591: The contraction associated with $\pm1/4+n\in \phi^{-1}f^{-1}(\bar x)$ 
592: is $z\mapsto \pm z/2+n$.
593: If $[r]_N=(k_1,k_2)\in\left(\phi^{-1}f^{-1}(\bar x)\right)^2$, 
594: then $K_r$ is the closed interval $[m_1,m_2]$, where
595: \begin{multline*}
596: m_1=\min\{2n_1,2n_2\},m_2=\max\{2n_1,2n_2\}
597: \text{ if }k_1=1/4+n_1,k_2=1/4+n_2, 
598: \end{multline*}
599: \begin{multline*}
600: m_1=\min\{2n_1,n_2-n_1\},m_2=\max\{2n_1,n_2-n_1\}\\
601: \text{ if }\{k_1,k_2\}=\{1/4+n_1,-1/4+n_2\},
602: \end{multline*}  
603: \begin{multline*}
604:  m_1=\min\{\frac{4n_1-2n_2}3,\frac{4n_2-2n_1}3\}, 
605: m_2=\max\{\frac{4n_1-2n_2}3,\frac{4n_2-2n_1}3\}\\
606: \text{ if }k_1=-1/4+n_1,k_2=-1/4+n_2.
607: \end{multline*}
608: The coding maps $\pi_r$ such that $[r]_N=(k_1,k_2)$ with $\{k_1,k_2\}
609: =\{1/4+n_1,-1/4+n_2\}$ ($|3n_1-n_2|=1$), for which the attractors $K_r$ 
610: are intervals of length one, are considered canonical. 
611: 
612: \subsection{}
613: Let $f(z)=-{(z-1)^2}/{4z}$.
614: (This map is conjugate to the map $z\mapsto (z-2)^2/z^2$, 
615: which is discussed in \cite{Be91}, \S4.3.)
616: Then $C=\{-1,1\},P=\{1,0,\infty\},\mathrm{AP}=\emptyset$ and 
617: $S=J=\hat{\mathbb C}$. 
618: The canonical ramification function is $\rho(0)=\rho(\infty)=4$, 
619: $\rho(1)=2$, and $\rho(z)=1$ otherwise.
620: Note that $\tilde S$ is Euclidean.
621: If $\phi:\mathbb C\to \hat{\mathbb C}$ is an elliptic function 
622: of order four with lattice $2\Gamma=\{2n+2mi\,|\,n,m\in\mathbb Z\}$ 
623: such that $\phi(iz)=\phi(z),
624: \phi(0)=\infty,\phi(1)=1,\phi(1+i)=0$, then $\phi(\alpha z+\beta)
625: =f(\phi(z))$ for $\alpha\in\{\pm1\pm i,\pm1\mp i\},
626: \beta\in2\Gamma$ and $\phi$ is a 
627: universal covering for the orbifold $(S,\rho)$.
628: Clearly, $\phi^{-1}(J)=\mathbb C$.
629: Note that $\phi(z)=a\wp(2z)^2$ satisfies the 
630: above condition, where $\wp$ is the 
631: Weierstrass elliptic function for the lattice $\Gamma=
632: \{n+mi\,|\,n,m\in\mathbb 
633: Z\}$ and $a$ is some constant.
634: It is easily seen that the map $f$ is a version of Latt\`es' example.
635: (For example, see \cite{Be91}, \S4.3 or \cite{Mi99}, \S7.
636: Using the addition formula, we have 
637: $\wp(\alpha z)^2=-(\wp(z)^2-g_2/4)^2/4\wp(z)^2$ for $\alpha\in
638: \{\pm1\pm i,\pm1\mp i\}$, where $g_2$ is a nonzero constant. 
639: The constant $a$ is equal to $4/g_2$.)
640: Fix basepoints $\bar x=-1$ and $\tilde x=1/2+i/2$.
641: Then $\phi^{-1}f^{-1}(\bar x)=\{s+n+mi\,|\,n,m\in\mathbb Z,s=1/2\text
642: { or }i/2\}$.
643: The contraction associated with $x\in\phi^{-1}f^{-1}(\bar x)$ is 
644: $$\begin{array}{ll}
645: z\mapsto (1-i)z/2+n+mi& \text{ if }x=1/2+n+mi\\ 
646: z\mapsto (1+i)z/2+n+mi& \text{ if }x=i/2+n+mi\\
647: z\mapsto (-1+i)z/2+n+mi& \text{ if }x=-1/2+n+mi\\
648: z\mapsto (-1-i)z/2+n+mi& \text{ if }x=-i/2+n+mi,\\
649: \end{array}$$
650: where $n+mi\in(1+i)\Gamma=\{n+mi\,|\,n+m\text{ is even}\}$.
651: The attractor $K_r$ is a compact set with integral Lebesgue measure. 
652: For example, in the case $[r]_N=(-i/2+1+i,-1/2+2)$, $K_r$ is the triangle 
653: with vertices $0,2,1+i$.
654: This case is considered as canonical.
655: In the case $[r]_N=(i/2,1/2+1+i)$, $K_r$ is the L\'evy Dragon (Figure 
656: \ref{fig:levy}).
657: Thus Tiling Theorem  gives another proof of the well-known 
658: fact that the L\'evy Dragon tiles $\mathbb R^2$ and has nonempty interior.
659: 
660: We can calculate the 2-dimensional Lebesgue measure $\mu(K_r)$ for 
661: $[r]_N=(\alpha_1/2+\beta_1,\alpha_2/2+\beta_2)$ 
662: ($\alpha_j\in\{\pm1,\pm i\},
663: \beta_j\in(1+i)\Gamma$) as follows. 
664: $$
665: \mu(K_r)=\left\{
666: \begin{array}{cl}
667: 2|\beta_2-\beta_1|^2&\text{ if }\alpha_1=\alpha_2\\
668: \left|\frac{\beta_2}{2-(1-i)\alpha_2}
669: -\frac{\beta_1}{2-(1-i)\alpha_1}\right|^2
670: &\text{ if }(\alpha_1,\alpha_2)=(1,i),(i,1)\\
671: 10\left|\frac{\beta_2}{2-(1-i)\alpha_2}
672: -\frac{\beta_1}{2-(1-i)\alpha_1}\right|^2
673: &\text{ if }(\alpha_1,\alpha_2)\in 
674: A_1\times A_2\text{ or }A_2\times A_1\\
675: 25\left|\frac{\beta_2}{2-(1-i)\alpha_2}
676: -\frac{\beta_1}{2-(1-i)\alpha_1}\right|^2
677: &\text{ if }(\alpha_1,\alpha_2)=(-1,-i),(-i,-1)
678: \end{array}
679: \right.
680: $$
681: where $A_1=\{1,i\},A_2=\{-1,-i\}$.
682: The details are left to the reader.
683: 
684: \begin{figure}[hbtp]
685: \centering
686: \includegraphics[height=8cm]{levy.ps}
687: \caption{The L\'evy Dragon.}
688: \label{fig:levy}
689: \end{figure}
690: 
691: \subsection{}\label{sec:3.1}
692: Let $f(z)=z^2-3$.
693: Then $C=\{0,\infty\}$, $P=\{-3,6,33,\cdots,\infty\}$, 
694: $\mathrm{AP}=\{\infty\}$, $S=\mathbb C$
695: ,
696:  and the Julia set $J$ is a Cantor set in the real axis.
697: The canonical ramification function is $\rho(z)=2$ if $z=-3,6,33,\dots$,
698: $\rho(z)=1$ otherwise.
699: Thus $\tilde S$ is hyperbolic (i.e. $S=\{|z|<1\}$), and for any radial,
700: the corresponding contractions $g_1,g_2$ are not invertible.
701: 
702: If we take radials $r_1$, $r_2$ and $r_3$ as in Figures \ref{fig:fig1}, 
703: \ref{fig:fig2} and \ref{fig:fig3}, then 
704:  $\pi_{r_1}$ is a homeomorphism, $\pi_{r_2}$ is exactly two-to-one, 
705: and $\pi_{r_3}$ is at most three-to-one respectively.
706: Moreover, $\#\pi_{r_3}^{-1}(x)=1$ for $\mu$-almost all $x$.
707: So, we say that $\pi_{r_3}$ is almost one-to-one.
708: See Section \ref{sec:5} for the proof.
709: The coding map $\pi_{r_1}$ is considered canonical.
710: 
711: \begin{figure}[hbtp]
712: \centering
713: \includegraphics[height=4cm]{fig1.eps}
714: \caption{The radial $r_1$.}
715: \label{fig:fig1}
716: \end{figure}
717: 
718: \begin{figure}[hbtp]
719: \centering
720: \includegraphics[height=4cm]{fig2.eps}
721: \caption{The radial $r_2$.}
722: \label{fig:fig2}
723: \end{figure}
724: 
725: \begin{figure}[hbtp]
726: \centering
727: \includegraphics[height=4cm]{fig3.eps}
728: \caption{The radial $r_3$.}
729: \label{fig:fig3}
730: \end{figure}
731: 
732: \section{Further setting}\label{sec:4}
733: 
734: In this section, we define an invariant subgroup and 
735: the equilibrium state (the Brolin-Lyubich measure).
736: 
737: Let $f$ be a subhyperbolic rational map of degree $d$, and 
738: $\rho=\rho_f$ the canonical ramification function.
739: Fix a basepoint $\bar x$.
740: 
741: \begin{defi}
742: For a subgroup $N\subset G=\pi_1(S',\bar x)$, we have a covering 
743: $$\phi_N:(S'_N,x^N)\to(S',\bar x)$$
744: with $\phi_{N*}\pi_1(S'_N,x^N)=N$.
745: If $N^\rho\subset N$, then 
746: we can extend $\phi_N$ to a branched covering 
747: $$\phi_N:(S_N,x^N)\to (S,\bar x),$$
748: where $S_N-\phi_N^{-1}(P)=S_N'$.
749: For $l\in \bar L(f,\bar x)$, we denote by $l^N$ the lift of $l$ to $S'_N$ 
750: (or $S_N$) with $l^N(0)=x^N$.
751: 
752: If $N'$ is a subgroup with $N'\subset N\subset G$, 
753: then there exists a unique covering 
754: $$\psi_{N',N}:(S'_{N'},x^{N'})\to (S'_N,x^N)$$
755: such that $\phi_N\circ\psi_{N',N}=\phi_{N'}$.
756: If $N^\rho\subset N'$, then $\psi_{N',N}$ is extended to a branched 
757: covering $\psi_{N',N}:S_{N'}\to S_{N}$. 
758: 
759: Suppose $N$ is a normal subgroup.
760: The group of deck transformations for $\phi_N$ is identified with 
761: $G/N$.
762: For $\gamma\in\Gamma(f,\bar x)$, we denote the quotient class for 
763: $[\gamma]$ by $[\gamma]_N$.
764: \end{defi}
765: 
766: \begin{defi}
767: Let $l\in\Lambda(f,\bar x)$ with terminal point $x$.
768: Consider a homomorphism $f_*:\pi_1(S'-f^{-1}(P),x)
769: \to G$ induced from $f$ and a homomorphism 
770: $l_{\#}^{-1}\iota_*:\pi_1(S'-f^{-1}(P),x)
771: \to G$ induced from the inclusion and the path $l^{-1}$.
772:  A subgroup $N\subset G$ is said to be {\em invariant} with respect 
773: to $l$ if $N\subset f_*(l_{\#}^{-1}\iota_*)^{-1}(N)$, or 
774: equivalently if $F_x(\gamma)$ is a closed curve and 
775:  $[lF_x(\gamma)l^{-1}]\in N$ for every $\gamma$ with $[\gamma]\in N$.
776: 
777: If $N$ and $N'$ are invariant with respect to $l$, then so is 
778: the subgroup generated by $N$ and $N'$.
779: If $N$ and $N'$ are invariant with respect to $l$ and $l'$ respectively, 
780: then $N\cap N'$ is invariant both with respect to $l$ and $l'$.
781: We denote by $N_l$ the maximal invariant subgroup with respect to $l$.
782: A subgroup $N$ is invariant with respect to a radial $r=(l_i)$ if 
783: $N$ is invariant with respect to all $l_i,i=1,2,\dots,d$.
784: The maximal invariant subgroup $N_r$ with respect to $r$ is equal to 
785: $\bigcap_{i=1}^d N_{l_i}$. 
786: It is evident that $N^\rho$ is invariant with respect to every 
787: radial for any ramification function $\rho$.
788: \end{defi}
789: 
790: \begin{pro}\label{pro:12}
791: Let $l\in\Lambda(f,\bar x)$ and let $N\subset G$ be a subgroup.
792: If $N$ is invariant with respect to $l$, then there exists a 
793: covering $g:S_N'\to S'_N-\phi_N^{-1}f^{-1}(P)$ such that 
794: $f\circ\phi_N\circ g=\phi_N$ and  $g(x^N)=l_N(1)$.
795: If $N^\rho\subset N$, then $g$ is extended to a branched covering 
796: $g:S_N\to S_N$.
797: \end{pro}
798: 
799: \begin{proof}
800: Let $[\gamma]\in N$.
801: Then $[lF_x(\gamma)l^{-1}]\in N$, and so $lF_x(\gamma)l^{-1}$ 
802: can be lifted to a closed curve in $S'_N-\phi_N^{-1}f^{-1}(P)$. 
803: This means $(f\circ \phi_N)_*\pi_1(S'_N-\phi_N^{-1}f^{-1}(P),l_N(1))
804: \supset N=\phi_{N*}\pi_1(S'_N,x^N)$.
805: Therefore we have a covering $g:S'_N\to S'_N-\phi_N^{-1}
806: f^{-1}(P)$ with $f\circ\phi_N\circ g=\phi_N$ 
807: and  $g(x^N)=l_N(1)$. 
808: If $N^\rho\subset N$,  we can extend the map $g$ on $S_N$.
809: \end{proof}
810: 
811: \begin{rem}
812: The covering $g$ above is contractive in the pullback metric 
813: $\phi_N^*\nu$,  where $\nu$ is the expanding metric for $f$. 
814: \end{rem}
815: 
816: Let $r=(l_i)$ be a radial, and $N$ an invariant subgroup with 
817: respect to $r$.
818: There exist 
819: contractions $g_1,g_2,\dots,g_d:S_N'\to S_N'$ such that
820: $f\circ\phi_N\circ g_i=\phi_N$ and $g_i(x^N)=l_i^N(1)$.
821: Similarly to Proposition \ref{pro:ss}, 
822: \begin{equation}
823:  l_w^N(1)=g_w(x^N).
824: \label{eq:4-1}
825: \end{equation}
826: For simplicity, we assume $N^\rho\subset N$.
827: 
828: \begin{defi}
829: We have a nonempty compact set $K=K^N=K_r^N\subset S_N$ with
830: $K=\bigcup_{i=1}^d g_i(K)$, and a coding map $\pi^N=\pi_r^N:\Sigma\to K$ 
831: with $\pi^N(i\omega)=g_i\pi^N(\omega)$.
832: We say $K$ is a {\em Julia tile} if $\phi_N(K)=J$.
833: \end{defi}
834: 
835: From (\ref{eq:4-1}), we have 
836: \begin{equation*}
837: \pi=\phi_N\circ\pi^N.
838: \label{eq:4-2}
839: \end{equation*}
840: 
841: 
842: \begin{pro}\label{pro:13}
843: Let $r$ be a radial.
844:  The maximal invariant subgroup $N_r$ coincides with the normal 
845: subgroup 
846: $$N_r'=\{[\gamma]\in G\,|\, F_{x_w}(\gamma)\text{ is a closed curve for 
847: any }w\in W\}.$$
848: \end{pro}
849: 
850: \begin{proof}
851:  If $F_{x_w}(\gamma)$ is a closed curve for any $w\in W$, then 
852: $F_{x_w}(l_{x_i}F_{x_i}(\gamma)l_{x_i}^{-1})
853: =l_{x_{wi}}F_{x_{wi}}(\gamma)l_{x_{wi}}^{-1}$ is a closed curve 
854: for any $i=1,2,\dots,d$ and any $w\in W$.
855: Thus $N_r'$ is invariant with respect to $r$. 
856: 
857: If $[\gamma]\in N_r$, then $F_{x_i}(\gamma)$ is a closed curve 
858: and $[l_iF_{x_i}(\gamma)l_i^{-1}]\in N_r$ for $i=1,2,\dots,d$. 
859: Inductively, we see that $F_{x_w}(\gamma)$ is a closed curve and 
860: $[l_wF_{x_w}(\gamma)l_w^{-1}]\in N_r$ for $w\in W$. 
861: Thus $[\gamma]\in N_r'$.
862: \end{proof}
863: 
864: \begin{defi}
865: We consider the monodromy actions $\eta_k$ of $G$ on $f^{-k}(\bar x)$ for 
866: $k=1,2,\dots$, that is, $\eta_k([\gamma])(x)
867: =x\cdot [\gamma]=F_x(\gamma)(1)$.
868: We set 
869: $$\begin{array}{ccl}
870: \hat N&=&\bigcap_{k=1}^\infty\ker\eta_k\\
871: &=&\{[\gamma]\in G\,|\,\text{$F_x(\gamma)$ is a closed curve for any 
872: $k\ge1$ and any $x\in f^{-k}(\bar x)$}\}.
873: \end{array}$$
874: The quotient group 
875: $$\hat G=G/\hat N$$
876: is called the {\em reduced fundamental group} of $f$.
877: \end{defi}
878: 
879: From Proposition \ref{pro:13}, we immediately have
880: 
881: \begin{pro}
882:  If $l\in\Lambda(f,\bar x)$, then $\hat N$ is invariant with respect to 
883: $l$.
884: If $r=(l_i)$ is a proper radial, then $\hat N=N_r$.
885: \end{pro}
886: 
887: 
888: \begin{pro}\label{pro:1}
889: Let $l,l'\in\Lambda(f,\bar x)$, and let $N$ be an invariant subgroup 
890: with respect to $l$ and $l'$.
891: We have two coverings $g,g':S'_N\to S_N-\phi_N^{-1}f^{-1}(P)$ 
892: corresponding to $l,l'$ 
893: respectively by Proposition \ref{pro:12}. 
894: Suppose there exists a deck transformation $t:S'_N\to S'_N$ for the
895:  covering $\phi_N$ such that 
896: $t(z)=z'$ and $g(z)=g'(z')$ for some $z,z'\in  S'_N$.
897: Then  $g'\circ t=g$. 
898: 
899: In particular, if  $N$ is an invariant normal subgroup, then for the  
900: two coverings $g,g':S'_N\to S_N-\phi_N^{-1}f^{-1}(P)$ 
901: corresponding to $l,l'$, there exists a deck transformation $t$ 
902: with $g'\circ t=g$. 
903: \end{pro}
904: 
905: \begin{proof}
906: Write $y=g(z)=g'(z')$ and $x=\phi_N(z)=\phi_N(z')$. 
907: Consider the induced homomorphisms 
908: $$\begin{array}{rccc}
909: {g}_*:&\pi_1(S'_N,z)&\to &\pi_1(S'_N-\phi_N^{-1}f^{-1}(P),y),\\
910: {g'}_*:&\pi_1(S'_N,z')&\to&\pi_1(S'_N-\phi_N^{-1}f^{-1}(P),y),\\
911:  \phi_{N*}:&\pi_1(S'_N,z)&\to &\pi_1(S',x),\\
912:  \phi_{N*}':&\pi_1(S'_N,z')&\to &\pi_1(S',x),\\
913: (f\circ \phi_N)_*:&\pi_1(S'_N-\phi_N^{-1}f^{-1}(P),y)&\to &\pi_1(S',x).\\
914: \end{array}$$
915: The existence of $t$ implies $\phi_{N*}\pi_1(S'_N,z)=
916: \phi_{N*}'\pi_1(S'_N,z')$.
917: By (\ref{eq:0}) and the injectivity of $(f\circ\phi)_*$, we have 
918:  ${g}_*\pi_1(S'_N,z)={g}'_*\pi_1(S'_N,z')$.
919: Therefore there exists a unique homeomorphism $t':S_N\to S_N$
920: such that $g'\circ t=g$ and $t'(z)=z'$.
921: By the uniqueness of $t$, we have $t=t'$. 
922: \end{proof}
923: 
924: 
925: \begin{defi}
926: A Borel measure $\mu$ defined as follows is called the 
927: {\em Brolin-Lyubich measure} for $f$.
928: Let 
929: $$\mu_k=d^{-k}\sum_{y\in f^{-k}(\bar x)}\delta_y,$$
930:  where $\delta_y$ is the point mass at $y$. 
931: We define $\mu$ as the weak$*$ limit of $\mu_k$, that is, 
932: $\int hd\mu=\lim_{k\to\infty}\int hd\mu_k$ for any continuous function 
933: $h$.
934: It is known that the limit $\mu$ exists for any rational map of the 
935: Riemann sphere and $\mu$ is the unique equilibrium state, 
936: hence $f$ is strongly mixing with respect 
937: to $\mu$  (see \cite{Lj83}).
938: In our case, since $f$ is expanding, the direct proof is not difficult.
939: (We can also construct $\mu$ via a  Markov partition. 
940: Cf. \cite{Bo75}, \cite{Ja68}, \cite{Ja69}, \cite{Gu70}.)
941: 
942: \begin{pro}\label{pro:19}
943: The measure $\mu$ is characterized as the unique Borel probability 
944: measure satisfying 
945: \begin{equation}
946: \int h(x)d\mu(x)=d^{-1}\int \sum_{y\in f^{-1}(x)}h(y)d\mu(x)
947: \label{eq:equil}
948: \end{equation}
949: for any continuous function $h$.
950: \end{pro}
951: 
952: \begin{proof}
953: The operator $\mu\mapsto L(\mu)$ defined by 
954: $$\int h(x)dL(\mu)(x)=d^{-1}\int \sum_{y\in f^{-1}(x)}h(y)d\mu(x)$$ 
955: is a contraction of the space of Borel probability measures having  
956: compact supports in $S$ with metric $D(\mu,\mu')=\sup_{h}
957: |\int hd\mu-\int hd\mu'|$, where the supremum runs over all functions
958:  $h$ with Lipschitz constant (which is taken with respect to the
959:  expanding metric on $S$) 
960: less than or equal to one.
961: Indeed, if the Lipschitz constant of $h(x)$ is less than $\alpha$, then 
962: the Lipschitz constant of $d^{-1}\sum_{y\in f^{-1}(x)}h(y)$ is less than 
963: $c\alpha$, where $c$ is the contraction constant (see the end of 
964: Definition \ref{def:3}).
965: \end{proof}
966: 
967: \begin{rem}
968: Lyubich \cite{Lj83} proved that  
969: (\ref{eq:equil}) holds for any rational map $f$ 
970: and any Borel function $h(x)$.
971: \end{rem}
972: 
973: 
974: Let $N$ be a subgroup of $G$.
975: A Borel measure $\mu_N$ on $S'_N$ (or $S_N$) is defined as the lift of 
976: $\mu$, that is, for a small Borel set $E\subset S'_N$ for which the 
977: restriction $\phi|E$ is injective, we set $\mu_N(E)=\mu(\phi(E))$. 
978: We write 
979: $$J_N=\phi_N^{-1}(J).$$
980: It is easily seen that $\mu$ (or $\mu_N$) is supported on $J$ 
981: (or $J_N$).
982: \end{defi}
983: 
984: Hence we have
985: 
986: \begin{pro}\label{pro:3}\
987: \begin{enumerate}
988:  \item \label{pro:3-2}
989: $\mu$ is an invariant measure (i.e. $\mu(E)=\mu(f^{-1}(E))$ 
990: for any Borel set $E\subset S$). 
991: \item $f$ is strongly mixing with respect to $\mu$ (i.e.
992: $\lim_{k\to\infty}\mu(A\cap f^{-k}(B))=\mu(A)\mu(B)$ for Borel sets 
993: $A,B\subset S$).  
994: \item \label{pro:3-2.5}
995: For a Borel set $E\subset S$, $d\cdot\mu(E)=\mu(f(E))$ provided 
996: $f:E\to f(E)$ is injective.
997: \item \label{pro:3-3} 
998: Let $1\le i\le d$.
999: For a Borel set $E\subset S_N'$, 
1000: $d\cdot\mu_N(g_i(E))=\mu_N(E)$ provided $g_i:E\to g_i(E)$ is 
1001: injective. In general, if $g_i:E\to g_i(E)$ is at most $n$-to-one,
1002: then $d\cdot\mu_N(g_i(E))\leq \mu_N(E)\le nd\cdot\mu_N(g_i(E))$.
1003: \end{enumerate} 
1004: \end{pro}
1005: 
1006: \begin{proof}
1007: We need to show only (\ref{pro:3-2.5}).
1008: (\ref{pro:3-3}) is an immediate consequence of (\ref{pro:3-2.5}).
1009: 
1010: Suppose $E$ is so small that there exists an open set $U\supset 
1011: \overline E$ with $f:U\to f(U)$ injective.
1012: Then for any continuous function $h$ with support included in $U$, 
1013: we have $\int hd\mu=d^{-1}\cdot\int h\circ(f|U)^{-1}d\mu$  
1014: by Proposition \ref{pro:19}.
1015: Thus $\mu(E)=d^{-1}\mu(f(E))$.
1016: In general, divide $E$ into small parts.
1017: \end{proof}
1018: 
1019: \section{Julia tiling}\label{sec:tile}
1020: 
1021: In this section, we prove Tiling Theorem and Multiplicity Theorem.
1022: Moreover, we give several necessary and sufficient conditions for 
1023: suitableness of a coding map.
1024: 
1025: Let $f$ be a subhyperbolic rational map of degree $d$.
1026: Let $r$ be a radial with basepoint $\bar x$, 
1027: and $N$ an invariant normal subgroup with 
1028: respect to $r$ such that $N^{\rho_f}\subset N$.
1029: Write $K=K^N_r$.
1030: 
1031: \begin{thm}\label{thm:2}
1032:  If $\pi(\Sigma)=J$, then there exists a subset $T$ of the group 
1033: of deck transformations $G/N$ 
1034: such that $\mu_N(t(K)\cap t'(K))=0$ for $t\ne t'$ and 
1035: $\bigcup_{t\in T}t(K)=J_N$.
1036: \end{thm}
1037: 
1038: \begin{proof}
1039: By (\ref{pro:3-3}) of Proposition \ref{pro:3} 
1040: and $K=\bigcup_{i=1}^dg_i(K)$, we have 
1041: $$\mu_N(K)\le\sum_{i=1}^d\mu_N(g_i(K))\leq\mu_N(K).$$
1042: Thus 
1043: \begin{equation}
1044:   \mu_N(K)=\sum_{i=1}^d\mu_N(g_i(K))=d\cdot\mu_N(g_i(K)),\ 
1045: i=1,2,\dots,d,\label{eq:2}
1046: \end{equation}
1047: \begin{equation}
1048:  \sum_{i\ne j}\mu_N(g_i(K)\cap g_j(K))=0,
1049: \text{ and }
1050: \sum_{i=1}^d\mu_N(\{x\in K\,|\,\#(g_i^{-1}(x)\cap K)\ge2\})=0.
1051: \label{eq:3}\end{equation}
1052: 
1053: There exists a subset $U\subset K$ open 
1054: in the relative topology of $J_N$. 
1055: Indeed,  $\phi_N(K)=\phi_N(\pi^N(\Sigma))=\pi(\Sigma)=J$ implies 
1056: $J_N=\bigcup_{t\in G/N}t(K)$.
1057: Since $G/N$ is countable, $K$ has a nonempty interior in the relative 
1058: topology by the Baire category theorem.
1059: Take $\omega=\omega_1\omega_2\dots\in\Sigma$ with $\pi^N(\omega)\in U$.
1060: If $k_0$ is sufficiently large, 
1061: we have $g_w(K)\subset U$ for  $w=\omega_1\omega_2\cdots
1062: \omega_{k_0}\in W_{k_0}$.  
1063: Note that $\pi^N(w^\infty)\in U$. 
1064: Therefore for any $x\in J_N$, there exists $k\ge0$ such that 
1065: $g_w^k(x)\in U$.
1066: Consequently, $\bigcup_{k=0}^\infty g_w^{-k}(K)=J_N$.
1067: 
1068: For $\mu_N$-almost all $x\in J_N$, there uniquely exists 
1069: $y_x\in K$ such that $g_u(y_x)=g_w^k(x)$ for some $k\ge0$ and 
1070: $u\in W_{kk_0}$.
1071: Indeed, let 
1072: $$E_k=\bigcup_{u\ne w\in W_k}(g_u(K)\cap g_w(K))\cup\bigcup_{w\in W_k}
1073: \{x\in K\,|\, \#(g_w^{-1}(x)\cap K)\ge 2\}$$
1074: and $E'=\bigcup_{w\in W}g_w^{-1}(\bigcup_{k>0}E_k)$.
1075: Each $g_w:K\to K$ is at most finite-to-one.
1076: Thus $\mu_N(E')=0$ by (\ref{pro:3-3}) of Proposition \ref{pro:3} 
1077: and (\ref{eq:3}).
1078: Let $x\in J_N-E'$ and let $k\ge0$ be the minimal integer such that
1079: $g_w^k(x)\in K$. 
1080: Since $g_w^k(x)\notin E_{kk_0}$, there  exists a unique $u\in W_{kk_0}$ 
1081: such that 
1082: $g_w^k(x)\in g_u(K)$ and $\#g_u^{-1}(g_w^k(x))\cap K=1$.
1083: If $k'>k$ and $g_{u'}(y_x)=g_w^{k'}(x)$ for $u'\in W_{k'k_0}$, 
1084: then necessarily $u'=w^{k'-k}u$, and so $g_u^{-1}(g_w^k(x))\cap K
1085: =g_{u'}^{-1}(g_w^{k'}(x))\cap K$.
1086: 
1087: By Proposition \ref{pro:1}, there exists a unique deck transformation 
1088: $t_x\in G/N$ such that $g_w^k\circ t_x=g_u$ and $t_x(y_x)=x$.
1089: Let $T=\{t_x\,|\,x\in J_N-E'\}$.
1090: Then $\bigcup_{t\in T}t(K)$ includes $J_N-E'$.
1091: Since $\bigcup_{t\in T}t(K)$ is closed, it is equal to $J_N$.
1092: By the uniqueness of $y_x$, we can see that $(t_x(K)\cap t_{x'}(K))-E'
1093: =\emptyset$ whenever $t_x\ne t_{x'}$.
1094: \end{proof}
1095: 
1096: \begin{co}
1097:  If $\pi(\Sigma)=J$, then $K$ has nonempty interior in the relative 
1098: topology of $J_N$.
1099: \end{co}
1100: \begin{proof}
1101:  We have already shown this statement in the beginning of the second 
1102: paragraph of the proof of Theorem \ref{thm:2}.
1103: \end{proof}
1104: 
1105: \begin{co}\label{co:18}
1106:  If $\pi(\Sigma)=J$, then $d\cdot\mu_N(g_i(A))=\mu_N(A),
1107: 1\le i\le d$ for any Borel set $A\subset K$.
1108: \end{co}
1109: 
1110: \begin{proof}
1111: Let $A\subset K$ be a Borel set.  
1112: Let $E=E_1$ be the set defined in the proof of Theorem \ref{thm:2}.
1113: Then $g_i:A-g_i^{-1}(E)\to g_i(A)-E$ is one-to-one.
1114: Therefore $\mu_N(g_i(A))=\mu_N(g_i(A)-E)
1115: =d^{-1}\mu_N(A-g_i^{-1}(E))=d^{-1}\mu_N(A)$ 
1116: by (\ref{pro:3-3}) of Proposition \ref{pro:3}.
1117: \end{proof}
1118: 
1119: \begin{defi}
1120:  We denote by $m$ the {\em identically distributed Bernoulli measure} 
1121: on $\Sigma$.
1122: Namely, $m(\Sigma(w))=d^{-k}$ for every $w\in W_k$.
1123: \end{defi}
1124: 
1125: \begin{pro}\label{pro:20}
1126: If $\pi(\Sigma)=J$, then the conditional measure $\mu_N|K$ is 
1127: the invariant measure for the iterated function system 
1128: $\mathcal I$, that is, 
1129: \begin{equation}
1130: \mu_N(A\cap K)=d^{-1}\sum_{i=1}^d\mu_N(g_i^{-1}(A)\cap K)
1131: \label{eq:4.3}
1132: \end{equation}
1133: for any Borel set $A$.
1134: In particular, we have $\mu_N|K=\pi_{N*}m$ 
1135: (i.e. $\mu_N(A)/\mu_N(K)=m(\pi_N^{-1}(A))$ 
1136: for any Borel set $A\subset K$).
1137: \end{pro}
1138: 
1139: \begin{proof}
1140: (\ref{eq:4.3}) is an immediate consequence of Corollary \ref{co:18}.
1141: It is easy to see that $\pi_{N*}m$ is also the invariant measure 
1142: for $\mathcal I$ with weight $(1/d,1/d,\dots,1/d)$.
1143: The uniqueness of the invariant measure (\cite{Hu81}) implies
1144:  $\mu_N|K=\pi_{N*}m$.
1145: \end{proof}
1146: 
1147: 
1148: 
1149: \begin{thm}\label{thm:3}
1150: The following are equivalent.
1151: \begin{enumerate}
1152:  \item \label{it:thm3-0}
1153: $\pi(\Sigma)=J$.
1154: \item \label{it:thm3-0.5}
1155: For distinct words $w,w'\in W$, $[l_w{l_{w'}}^{-1}]$ is nontrivial, and 
1156: $\pi(\Sigma)\not\subset P$.
1157: \item \label{it:thm3-1}
1158: For distinct words $w,w'\in W$, $[l_w{l_{w'}}^{-1}]\notin N_r$.
1159: \item \label{it:thm3-3}
1160: $\lim_{k\to\infty}(\#\{g_w(x^N)\,|\,w\in W_k\})^{1/k}=d$ 
1161: for $N=\{1\}$, and $\pi(\Sigma)\not\subset P$.
1162: \item \label{it:thm3-3.5}
1163: $\lim_{k\to\infty}(\#\{g_w(x^N)\,|\,w\in W_k\})^{1/k}=d$ 
1164: for $N=N_r$.
1165: \item \label{it:thm3-4}
1166: For any $x\in J$, $\pi^{-1}(x)$ is finite.
1167: \item \label{it:thm3-5}
1168: $m(\{\omega\in\Sigma\,|\,\pi^{-1}\pi(\omega)\text{ 
1169: is at most countable}\})>0$.
1170: \item \label{it:thm3-6}
1171: $\mu(\pi(\Sigma))>0$.
1172: \end{enumerate} 
1173: \end{thm}
1174: 
1175: \begin{proof}
1176:  The implications (\ref{it:thm3-0})$\Rightarrow$(\ref{it:thm3-6})
1177: and (\ref{it:thm3-4})$\Rightarrow$(\ref{it:thm3-5}) 
1178: are trivial.
1179: 
1180: (\ref{it:thm3-0.5})$\Rightarrow$(\ref{it:thm3-3}),
1181: (\ref{it:thm3-1})$\Rightarrow$(\ref{it:thm3-3.5}):
1182: (\ref{it:thm3-0.5}) or (\ref{it:thm3-1}) implies 
1183: $\#\{g_w(x^N)\,|\,w\in W_k\}=d^k$. 
1184: Recall (\ref{eq:4-1}).
1185: 
1186: (\ref{it:thm3-3})$\Rightarrow$(\ref{it:thm3-0.5}),
1187: (\ref{it:thm3-3.5})$\Rightarrow$(\ref{it:thm3-1}):
1188: If $[l_wl_{w'}^{-1}]\in N$ for some $w,w'\in W_n$, then $g_w=g_{w'}$.
1189: Thus $\#\{g_u(x^N)\,|\,u\in W_{kn}\}\le (d^n-1)^{k}$.
1190: 
1191: (\ref{it:thm3-6})$\Rightarrow$(\ref{it:thm3-0}):
1192: Since $f\pi(\Sigma)=\pi(\Sigma)$, we have $\mu(\pi(\Sigma))=1$ 
1193: by the ergodicity.
1194: Thus (\ref{it:thm3-0}) follows from the compactness of $\pi(\Sigma)$.
1195: 
1196: (\ref{it:thm3-1})$\Rightarrow$(\ref{it:thm3-6}):
1197: Since $|l_w|$ is bounded for $w\in W$, it follows from 
1198: (\ref{it:thm3-1}) that there exists $n>0$ such that 
1199: \begin{equation}
1200: \#\{w'\,|\,
1201: x_{w'}=x_w\}\le n
1202: \label{eq:n}
1203: \end{equation}
1204:  for any $w\in W$.
1205: 
1206: Let $V$ be an open set including $\pi(\Sigma)$.
1207: Take an open set $V'$ such that $\pi(\Sigma)\subset V'$ and 
1208: $\overline {V'}\subset V$.
1209: Then there exists $k_0>0$ such that if $k>k_0$, then $x_w\in 
1210: V'$ for every $w\in W_k $.
1211: Hence $$d^k/n\le\#\{x_w\,|\,w\in W_k\}\cap V'\le\#f^{-1}(\bar x)\cap V'.$$
1212: Therefore $1/n\leq\mu(V)$, and so $1/n\leq\mu(\pi(\Sigma))$.
1213: 
1214: (\ref{it:thm3-1})$\Rightarrow$(\ref{it:thm3-0.5}):
1215: Since (\ref{it:thm3-1})$\Rightarrow$(\ref{it:thm3-0}), 
1216: (\ref{it:thm3-1}) implies $\pi(\Sigma)\not\subset P$.
1217: 
1218: 
1219: (\ref{it:thm3-5})$\Rightarrow$(\ref{it:thm3-1}):
1220: Suppose there exist distinct $w,w'\in W$ such that $[l_w{l_{w'}}^{-1}]
1221: \in N_r$.
1222: Set 
1223: $$\Omega=\{\omega\in\Sigma\,|\,\sigma^k\omega\in \Sigma(w)\text{ 
1224: for inifinitely many $k$}\}.$$
1225: Then $m(\Omega)=1$.
1226: For any $\omega=u_1wu_2wu_3\cdots\in\Omega$, 
1227: $\pi^{-1}\pi(\omega)$ includes an uncountable set 
1228: $\{u_1w_1u_2w_2u_3\cdots\,|\,w_i=w\text{ or }w'\}$.
1229: Indeed, note that if $[l_v{l_{v'}}^{-1}],[l_u{l_{u'}}^{-1}]\in  N_r$, 
1230: then $[l_{vu}{l_{v'u'}}^{-1}]\in N_r$.
1231: Thus $x_{u_1wu_2w\cdots u_kw}=x_{u_1w_1u_2w_2\cdots u_kw_k}$ if 
1232: $w_i=w\text{ or } w'$. 
1233: Hence $\pi(\{u_1w_1u_2w_2u_3\cdots\,|\,w_i=w\text{ or }w'\})=\{
1234: \pi(\omega)\}$.
1235: 
1236: (\ref{it:thm3-0.5})$\Rightarrow$(\ref{it:thm3-1}):
1237: Suppose $\pi(\Sigma)\not\subset P$ and 
1238: there exist distinct $w,w'\in W$ such that $[l_w{l_{w'}}^{-1}]
1239: \in  N_r$.
1240: If we take a large $k_0$, then $|F_{x_u}(l_{w}{l_{w'}}^{-1})|$ is so 
1241: small for $u\in W_k,k\ge k_0$ 
1242: that $F_{x_u}(l_{w}{l_{w'}}^{-1})$ is either homotopically trivial or 
1243:  winding around some point of $P$ several times in $S'$.
1244: Since $\pi(\Sigma)\not\subset P$, there exists a word $u$ such that 
1245: $F_{x_u}(l_{w}{l_{w'}}^{-1})$ is homotopically trivial.
1246: Then $[l_{uw}{l_{uw'}}^{-1}]$ is trivial.
1247: 
1248: (\ref{it:thm3-6})$\Rightarrow$(\ref{it:thm3-1}):
1249: Suppose there exists distinct $w,w'\in W$ such that $[l_w{l_{w'}}^{-1}]
1250: \in N_r$.
1251: Then $g_w=g_{w'}$ by (\ref{eq:4-1}). 
1252: From (\ref{eq:3}), we have  $\mu_N(g_w(K))=0$, and hence $\mu_N(K)=0$ 
1253: by (\ref{eq:2}).
1254: Therefore $\mu(\pi(\Sigma))=0$.
1255: 
1256: (\ref{it:thm3-1})$\Rightarrow$(\ref{it:thm3-4}):
1257: Let $p\in J$.
1258: Since $|l_\omega|$ is bounded for $\omega\in\Sigma$, 
1259: $L(p)=\{l_\omega\,|\,\omega\in\pi^{-1}(p)\}$ is divided into 
1260: a bounded number of  homotopy classes mod $N_r$, that is, 
1261: $$L(p)=L(p)_1\sqcup L(p)_2\sqcup\cdots \sqcup L(p)_{m(p)}\quad
1262: \left(l_\omega,l_{\omega'}\in L(p)_i\iff [l_\omega{l_{\omega'}}^{-1}]
1263: \in N_r\right)$$
1264: with $m(p)\le M$ for some $M>0$ independent of $p$.
1265: 
1266: Set
1267: $$\begin{array}{ccl}
1268:  A(p,k)&=&
1269:   \{w\in W_k\,|\,\Sigma(w)\cap\pi^{-1}(p)\ne\emptyset\}\\
1270: &=&\{w\in W_k\,|\, \text{ there exists $\omega\in \pi^{-1}f^k(p)$ 
1271: such that $\pi(w\omega)=p$}\},\\
1272: B(p,k,\omega)&=&\{x_w\,|\,w\in W_k,\pi(w\omega)=p\}.
1273: \end{array}
1274: $$
1275: Then $\#B(p,k,\omega)$ is equal to or less than $\deg_pf^k$, 
1276: the local degree of $f^k$ at $p$.
1277: Indeed, if $\pi(\omega)=f^k(p)$, then we have exactly $\deg_pf^k$ lifts 
1278: of $l_\omega$ by $f^{k}$ with terminal point $p$.
1279: Thus there exists $b>0$ independent of $p$, $k$ and $\omega$ 
1280: such that $\#B(p,k,\omega)<b$.
1281: It is easily seen that $B(p,k,\omega)=B(p,k,\omega')$ whenever 
1282: $[l_\omega{l_{\omega'}}^{-1}]\in N_r$.
1283: Therefore choosing $\omega_i$ so that $l_{\omega_i}\in L(f^k(p))_i$, 
1284: we have 
1285: $$\begin{array}{ccl}
1286: \#A(p,k)&\le &n\cdot\#\{x_w\,|\,w\in A(p,k)\}\\
1287: &=&n\cdot\#\bigcup_{\omega\in\pi^{-1}f^k(p)}B(p,k,\omega)\\
1288: &=&n\cdot\#\bigcup_{i=1}^{m(f^k(p))}B(p,k,\omega_i)\\
1289: &\le&nMb.
1290: \end{array}$$
1291: for any $k>0$, where $n$ is defined in (\ref{eq:n}).
1292: Hence 
1293: \begin{equation}\#\pi^{-1}(p)\le nMb,
1294: \label{eq:l}
1295: \end{equation}
1296: since $\pi^{-1}(p)= \bigcap_{k=1}^\infty
1297: \bigcup_{w\in A(p,k)}\Sigma(w)$.
1298: \end{proof}
1299: 
1300: 
1301: 
1302: 
1303: \begin{thm}\label{thm:21}
1304:  There exists an integer $n\ge 0$ such that for any invariant subgroup 
1305: $N\subset G$ with respect to $r$,
1306: \begin{enumerate}
1307: \item $\#\pi^{-1}(p)=\#\phi_N^{-1}(p)\cap K=n$ for $\mu$-almost 
1308: all $p\in J$, 
1309: \item $\#\pi^{-1}(p)\ge\#\phi_N^{-1}(p)\cap K\ge n$ 
1310: for any $p\in J-P$, and
1311: \item $\mu_N(\phi_N^{-1}(A)\cap K)=n\mu(A)$ for any Borel set $A\subset 
1312: \hat{\mathbb C}$.
1313: In particular, $\mu_N(K)=n$.
1314: \end{enumerate}
1315: \end{thm}
1316: 
1317: \begin{proof}
1318: First remark that $\pi=\phi_N\circ\pi^N$ and $\phi_N=\psi_{N,N_r}
1319: \circ\phi_{N_r}$ imply $\#\pi^{-1}(p)\ge\#
1320: \phi_N^{-1}(p)\cap K^N\ge\#\phi_{N_r}^{-1}(p)\cap K^{N_r}$ 
1321: for every $p\in J$.
1322: Let $E_k\subset S_{N_r}$ be the set as in the proof of Theorem 
1323: \ref{thm:2}.
1324: Then $D=\bigcup_{k=1}^\infty E_k$ is a null set.
1325: If $x\in K^{N_r}-D$, then $\#({\pi^{N_r}})^{-1}(x)=1$.
1326: Therefore $\#\pi^{-1}(p)=
1327: \#\phi_{N_r}^{-1}(p)\cap K^{N_r}$ for every $p\in J-\phi_{N_r}(D)$.
1328: 
1329: If $\pi(\Sigma)\ne J$, then $\mu(\pi(\Sigma))=0$ by Theorem 
1330: \ref{thm:3}, so the conditions are satisfied for $n=0$.
1331: 
1332: Suppose $\pi(\Sigma)=J$.
1333: We show that  $h(x)=\#\phi_{N_r}^{-1}(x)\cap K$ is a Borel function.
1334: To this end, let $A_{k,\epsilon}$ be the set of $x\in J$ 
1335: such that for some $z_i\in K,i=1,2,\dots,k$, 
1336: $\phi_{N_r}(z_i)=x,1\le i\le k$ and
1337: the distance between $z_i$ and $z_j$ is  equal to or bigger than 
1338: $\epsilon$ for $0\le i\ne j\le k$.
1339: Then  $A_{k,\epsilon}$ is closed, and so  
1340:  $\{x\,|\,h(x)\ge k\}=\bigcup_{n=1}^\infty A_{k-1,1/n}$ 
1341: is a Borel set for $k\ge 2$.
1342: 
1343: For a Borel set $E\subset J$,
1344: using $h(x)=d^{-1}\sum_{y\in f^{-1}(x)}h(y)$ and $\int_Eh(x)d\mu(x)=
1345: \int_{f^{-1}(E)}h\circ f(y)d\mu(y)$ (substitute $h$ of (\ref{eq:equil}) 
1346: by $1_{f^{-1}(E)}\cdot h\circ f$), we have
1347: $$\int_Eh(x)d\mu(x) = \int_{f^{-1}(E)}h(y)d\mu(y).$$
1348: Hence for any $E$ and any $k>0$, we have 
1349: $\int_Eh(x)d\mu(x)=\int_{f^{-k}(E)}d\mu(x)$, 
1350: which converges to $\mu(E)\int_Jh(x)d\mu(x)$ as $k\to\infty$ 
1351: by the strong mixing condition.
1352: Therefore $h(x)$ is constant almost everywhere.
1353: Thus (1) is verified.
1354: By the definition of $\mu_N$, we see that $\mu_N(\phi_N^{-1}(A)
1355: \cap K)=n\mu(A)$ for any Borel set $A$.
1356: 
1357: If $z_i,z_i'\in K,i=1,2,\dots$ satisfy $\lim_{i\to\infty}z_i=
1358: \lim_{i\to\infty}z_i'=z$, $z_i\ne z_i'$, and $\phi_N(z_i)=\phi_N(z_i')$, 
1359: then $\phi_N(z)\in P$.
1360: Since $\{x\,|\,\#\phi_N^{-1}(x)\cap K=n\}$ is dense in $J$, we have 
1361: $\#\phi_N^{-1}(x)\cap K\ge n$ for every $x\in J-P$.
1362: In general, $\#\phi_N^{-1}(x)\cap K\ge \max\{n-\max_{y\in \phi_N^{-1}(x)}
1363: \deg_{y}\phi_N+1,1\}$ for every $x\in J$.
1364: \end{proof}
1365: 
1366: 
1367: \begin{defi}
1368:  We call the number $n$ above the {\em multiplicity} of $\pi_r$ 
1369: and denote by $n_r$ or $\mathrm{mul}(\pi_r)$.
1370: \end{defi}
1371: 
1372: \begin{co}
1373:  If $\pi(\Sigma)=J$, then $\mu=\pi_*(m)$.
1374: \end{co}
1375: 
1376: \begin{proof}
1377: By (3) of Theorem \ref{thm:21} and Proposition \ref{pro:20},
1378: $m(\pi^{-1}(A))=\tilde\mu(\phi^{-1}(A)\cap K)/{n_r}=\mu(A)$ for 
1379: any Borel set $A\subset\hat{\mathbb C}$. 
1380: \end{proof}
1381: 
1382: 
1383: \begin{pro}
1384: If $\pi(\Sigma)=J$, then  the multiplicity of $\pi_r$ is equal to or 
1385: bigger than $$\max_{x\in\{x_1,x_2,\dots,x_d\}}\#\{i\,|\,x=x_i\}.$$
1386: In particular, if $r$ is not proper, then $n_r\ne 1$.
1387: \end{pro}
1388: 
1389: \begin{proof}
1390: Suppose $x_1=x_2=\cdots =x_s$.
1391: Then $\pi(i\omega)=\pi(j\omega)$ for any $\omega\in\Sigma$ whenever 
1392: $i,j\in\{1,2,\dots,s\}$.
1393: Thus $\#\pi^{-1}\pi(i\omega)\ge s$ for $1\le i\le s$ and $\omega\in
1394: \Sigma$.
1395: Since $\mu(\pi(\Sigma(i)))\ge m(\Sigma(i))=d^{-1}>0$, 
1396: the multiplicity is equal to or bigger than $s$.
1397: \end{proof}
1398: 
1399: \section{Structure of $\mathrm{Cod}(f)$}\label{sec:st}
1400: 
1401: In this section, we prove Structure Theorem, and show that the monoid 
1402: $\mathcal A(f)$ naturally acts on $\mathrm{Cod}(f)$.
1403: The structures of $\mathrm{Cod}(f)$ for a couple of examples are 
1404: closely investigated.
1405: 
1406: \begin{defi}
1407: Let $N\subset G$ be a subgroup.
1408: We say that $l$ and $l'\in \Lambda(f,\bar x)$ are 
1409: {\em homotopic modulo $N$ with basepoint held fixed} 
1410: if $l(1)=l'(1)$ and $[l_i{l_i'}^{-1}]\in N$.
1411: We denote by $\Lambda_N(f,\bar x)$ the set of homotopy classes 
1412:  modulo $N$ with basepoint $\bar x$ held fixed.
1413: Then $\Lambda_N(f,\bar x)$ is identified with 
1414: $\phi_N^{-1}f^{-1}(\bar x)$, and with $\{g:S'_N\to S'_N\,|\,
1415: f\circ\phi_N\circ g=\phi_N\}$ as well.
1416: The natural projection from $\Lambda(f,\bar x)$ to $\Lambda_N(f,\bar x)$ 
1417: is denoted by 
1418: $$l\mapsto [l]_N.$$
1419: 
1420: We say that two radials $r=(l_i)$ and $r'=(l_i')\in\mathrm{Rad}
1421: (f,\bar x)=\Lambda(f,\bar x)^d$ are 
1422: {\em  homotopic modulo $N$ with basepoint held fixed} 
1423: if $[l_i]_N=[l_i']_N$ for $i=1,2,\dots,d$.
1424: We write $$\mathrm{Rad}_N(f,\bar x)=\Lambda_N(f,\bar x)^d\text{ and }
1425: [r]_N=([l_i]_N)_i.$$
1426: \end{defi}
1427: 
1428: \begin{defi}\label{def:6}
1429:  We say that two radials $r\in\mathrm{Rad}(f,\bar x)$ and 
1430: $r'\in\mathrm{Rad}
1431: (f,\bar x')$ are {\em freely homotopic} if there exists a 
1432: homotopy 
1433: $H:Q\times[0,1]\to S'$ between $r$ and $r'$ such that $H(\cdot,t):
1434: Q\to S'$ is a radial for every $0\le t\le1$. 
1435: \end{defi}
1436: 
1437: Let 
1438: $$Y=\{\gamma:[0,1]\to S'\,|\,\gamma(0)=\bar x,\gamma(1)=\bar x'\}
1439: \subset L.$$
1440: We define an operation of $\gamma\in Y$ from $\Lambda(f,\bar x)$ to 
1441: $\Lambda(f,\bar x')$ by 
1442: $$\gamma\cdot l=\gamma^{-1} l F_{x_i}(\gamma).$$ 
1443: For a subgroup $N\subset G$, the operation $\gamma:\Lambda(f,\bar x)
1444: \to\Lambda(f,\bar x')$ naturally descends to 
1445: $\gamma:\Lambda_N(f,\bar x)\to\Lambda_{\gamma_{\#}^{-1}(N)}(f,\bar x')$. 
1446: Suppose $N'$ is invariant with respect to $l$ and $\gamma,\gamma'\in Y$ 
1447: satisfy $[\gamma{\gamma'}^{-1}]\in N'$.
1448: Then for $N\supset N'$, $\gamma\cdot[l]_{N}=\gamma'\cdot[l]_{N}$.
1449: In particular, in the case $\bar x=\bar x'$,  the operation 
1450: \begin{equation}
1451: [\gamma]_{\hat N}:\Lambda_{N}(f,\bar x)\to\Lambda_{\gamma^{-1}_{\#}
1452: (N)}(f,\bar x)\label{eq:6.1}
1453: \end{equation} 
1454: is well-defined for every $[\gamma]_{\hat N}\in \hat G=G/\hat N$.
1455: Identifying $\hat G$ with the group of deck transformations of 
1456: $\phi_{\hat N}:S_{\hat N}\to S$, we write the 
1457: action $t:\Lambda_N(f,\bar x)\to\Lambda_{tN}(f,\bar x)$  by 
1458: $$t\cdot g_i=tg_it^{-1}$$
1459: for $t\in \hat G$, 
1460: where we use the identification 
1461: $\Lambda_N(f,\bar x)=\{g:S_N\to S_N\,|\,f\circ\phi_N\circ g=\phi_N\}$.
1462: Clearly, the operation of $\gamma\in Y$ from $\mathrm{Rad}(f,\bar x)$ to 
1463: $\mathrm{Rad}(f,\bar x')$ is defined diagonally.
1464: From (\ref{eq:6.1})
1465: \begin{equation}
1466: [\gamma]_{\hat N}:\{[r]_{N_r}\,|\,r\in\mathrm{Rad}(f,\bar x)\}\to
1467: \{[r]_{N_r}\,|\,r\in\mathrm{Rad}(f,\bar x)\}
1468: \label{eq:6.2}
1469: \end{equation} 
1470: is well-defined for every $[\gamma]_{\hat N}\in \hat G$.
1471: 
1472: Two radials $r\in\mathrm{Rad}(f,\bar x)$ and $r'\in\mathrm{Rad}
1473: (f,\bar x')$ are freely homotopic  if and only if 
1474: there exists $\gamma\in Y$ such that $\gamma\cdot [r]_e=[r']_{e'}$, where 
1475: $e$ and $e'$ are the trivial subgroups.
1476: Thus we have a generalization of Definition \ref{def:6}.
1477: 
1478: \begin{defi}\label{def:19}
1479: Let $N\subset G$ be a subgroup.
1480: We say that $r$ and $r'$ are {\em freely homotopic modulo $N$} 
1481: if there exists 
1482: $\gamma\in Y$ such that $\gamma\cdot[r]_N=[r']_{\gamma_{\#}^{-1}(N)}$.
1483: \end{defi}
1484: 
1485: 
1486: \begin{thm}\label{thm:24}
1487: Let $r=(l_i)\in\mathrm{Rad}(f,\bar x)$ and $r'=(l_i')\in
1488: \mathrm{Rad}(f,\bar x')$.
1489: Then $\pi_r=\pi_{r'}$ if and only if $r$ and $r'$ are freely homotopic 
1490: modulo $N_r$. 
1491: \end{thm}
1492: 
1493: \begin{proof}
1494: For $r'$, the notation $x_w'$, $l'_w$, $l'_\omega$ and $F'_x(\cdot)$ 
1495: are defined in a trivial way.
1496: 
1497: Suppose $\pi=\pi_r=\pi_{r'}$.
1498: For $\omega\in\Sigma$, write $\gamma_\omega=l_\omega{l'_\omega}^{-1}$. 
1499: Remark that 
1500: \begin{equation}
1501: F_{x_w}(\gamma_\omega)=F_{x_w}(l_\omega)
1502: F'_{x'_w}(l'_\omega)^{-1}
1503: \label{eq:6-2}
1504: \end{equation}
1505:  for $w\in W$ and $\omega\in\Sigma$ with 
1506: $\pi(\omega)\notin P$, and 
1507: \begin{equation}
1508: l_wF_{x_w}(l_\omega)F'_{x'_w}(l'_\omega)^{-1}{l'_w}^{-1}
1509: =\gamma_{w\omega}
1510: \label{eq:6-3}
1511: \end{equation}
1512:  for $w\in W$ and $\omega\in\Sigma$.
1513: Since the curve $F_{x_w}(l_\omega)F'_{x'_w}(l'_\omega)^{-1}$ 
1514: joins $x_w$ and $x'_w$, 
1515: the curve $F_{x_w}(l_\omega)F'_{x'_w}(l'_\omega)^{-1}F'_{x'_w}
1516: (l'_{\omega'})F_{x_w}(l_{\omega'})^{-1}$ is closed for 
1517: any $w\in W$ and $\omega,\omega'\in\Sigma$.
1518: It follows from this that 
1519: $[\gamma_\omega{\gamma_{\omega'}}^{-1}]\in N_r$ provided 
1520: $\pi(\omega),\pi(\omega')\notin P$.
1521: 
1522: First we assume $\pi(\Sigma)\not\subset P$.
1523: By (\ref{eq:6-2}) and (\ref{eq:6-3}),
1524:  $\gamma_\omega^{-1}l_iF_{x_i}(\gamma_\omega)$ and 
1525: $\gamma_\omega^{-1}\gamma_{i\omega}{l_i'}$ are homotopic in $S'$ 
1526: with endpoints held fixed for $i=1,2,\dots,d$ 
1527: whenever $\pi(\omega)\notin P$. 
1528: Hence, since $[\gamma_\omega^{-1}\gamma_{i\omega}]\in 
1529: \gamma^{-1}_{\omega\#}N_r$, we have
1530:  $\gamma_\omega\cdot[r]_{N_r}=[r']_{\gamma_{\omega\#}^{-1}(N_r)}$.
1531: If $\pi(\Sigma)\subset P$, then $\pi(\Sigma)$ consists of one fixed
1532:  point (see Remark \ref{rem:5}), say $p$.
1533: Note that $f$ is one-to-one near $p$.
1534: Therefore if we modify $\gamma_\omega$ near $p$ into $\gamma'_\omega$ 
1535: avoiding $P$, 
1536: then $F_{x_w}(\gamma_\omega')$ coincides with $F_{x_w}(l_\omega)F'_{x'_w}
1537: (l'_\omega)^{-1}$ 
1538: except near $p$.
1539: Hence we use the same discussion as above to conclude that $r$ and $r'$ 
1540: are freely homotopic modulo $N_r$.
1541: 
1542: Suppose $r$ and $r'$ are freely homotopic modulo $N_r$.
1543: Then there exists $\gamma\in Y$ such that 
1544: $[\gamma l'_iF_{x_i}(\gamma)^{-1}
1545: {l_i}^{-1}]\in {N_r}$ for $i=1,2,\dots,d$.
1546: Let $x'\in\phi_{N_r}^{-1}(\bar x')$ be the terminal point of 
1547: the lift $\gamma^{N_r}$ of $\gamma$ to $S_{N_r}$ whose initial point 
1548: is $x^N$.
1549: We have the contractions $g'_1,g'_2,\dots,g'_d:S_{N_r}\to S_{N_r}$  
1550: corresponding to $r'$ with respect to the basepoints 
1551: $\bar x'$ and $x'$.
1552: Let $l_i^{'N_r}$ denote the lift of $l_i'$ to $S_{N_r}$ whose initial 
1553: point is $x'$, and $l_i^{N_r}\gamma_0^{N_r}$ the lift of 
1554: $l_iF_{x_i}(\gamma)$ to $S_{N_r}$ whose initial point is $x^{N_r}$.
1555: Then the terminal points of $l_i^{N_r},l_i^{'N_r}$ are 
1556: $g_i(x^{N_r}),g_i'(x')$ respectively.
1557: By assumption, the terminal point of $\gamma^{N_r}l_i^{'N_r}$ coincides 
1558: with that of $l_i^{N_r}\gamma_0^{N_r}$.
1559: Thus $\gamma_0^{N_r}$ joins $g_i(x^{N_r})$ and $g_i'(x')$.
1560: Since $\gamma_0^{N_r}$ is a lift of $\gamma$ by $f\circ\phi_{N_r}$, 
1561: we have $g_i(\gamma^{N_r})=\gamma_0^{N_r}=g_i'(\gamma^{N_r})$.
1562: Hence $g_i=g_i'$.
1563: Consequently, $\pi_r^{N_r}=\pi_{r'}^{N_r}$, and so 
1564: $\pi_r=\phi_{N_r}\circ\pi_r^{N_r}=\phi_{N_r}\circ\pi_{r'}^{N_r}=\pi_{r'}$.
1565: \end{proof}
1566: 
1567: \begin{co}\label{co:34-1}
1568: $$\mathrm{Cod}(f)\approx\{[r]_{N_r}\,|\,
1569: r\in\mathrm{Rad}(f,\bar x)\}/\hat G,$$
1570: where the action of $\hat G$ is defined in (\ref{eq:6.2}).
1571: \end{co}
1572: 
1573: 
1574: \begin{co}\label{co:35}
1575:  If $\pi_r=\pi_{r'}$, then $N_r$ and $N_{r'}$ are conjugate 
1576: (i.e. $N_{r'}=\gamma^{-1}_{\#}(N_r)$ for some $\gamma\in Y$).
1577: \end{co}
1578: 
1579: 
1580: 
1581: \begin{exa}\ 
1582: \begin{enumerate}
1583:  \item $f(z)=z^d$.
1584: Since $N^{\rho_f}$ is trivial, the fundamental group $G^{\rho_f}$ is equal
1585: to $G=\mathbb Z$.
1586: For any curve $l\in\Lambda(f,\bar x)$, the maximal invariant subgroup 
1587: $N_l$ is trivial.
1588: Thus $\hat G=G^{\rho_f}$ and 
1589: $\mathrm{Cod}(f)\approx (\phi^{-1}f^{-1}(\bar x))^d/\hat G=
1590: \mathbb Z^d/\mathbb Z$, where $n/d\in\phi^{-1}f^{-1}(\bar x)$ is 
1591: identified with $n\in\mathbb Z$, and $\mathbb Z$ acts on $\mathbb Z^d$ by 
1592: $$n\cdot(n_1,n_2,\dots,n_d)=(n_1+(d-1)n,n_2+(d-1)n,\dots,n_d+(d-1)n).$$
1593: Thus $$\mathrm{Cod}(f)\approx\{[n_1,n_2,\dots,n_d]\,:\, n_i\in\mathbb Z,
1594: 0\le n_1\le d-2\},$$ where $[n_1,\dots]$ denotes the equivalence class 
1595: including $(n_1,\dots)$.
1596: 
1597: 
1598: \item
1599: $f(z)=z^2-2$.
1600: The fundamental group $G^{\rho_f}$ is equal to $\mathrm{Iso}(2\mathbb Z)=
1601: \{x\mapsto ax+2n\,|\,a=\pm1,n\in\mathbb Z\}$.
1602: For $l\in\Lambda(f,\bar x)$, 
1603: let  $l_\infty=\phi(x)$ be the fixed point of $f$ such that 
1604: $g(x)=x$ for the contraction $g$ with respect to $l$.
1605: If $l_\infty=-1$, then $N_l=N^{\rho_f}$; 
1606: if $l_\infty=2$, then $N_l/N^{\rho_f}=\{\mathrm{id},x\mapsto -x+4n\}
1607: \subset\mathrm{Iso}(2\mathbb Z)$ for some $n\in\mathbb Z$.
1608: Therefore if $\pi_r(\Sigma)\ne\{2\}$, $N_r=N^{\rho_f}$; otherwise, 
1609: $N_r/N^{\rho_f}=\{\mathrm{id},x\mapsto -x+4n\}$.
1610: In the latter case, the branched covering $\phi_{N_r}:S_{N_r}\to S$ is 
1611: given by $\phi_{N_r}(z)=\phi(\sqrt z)=2\cos 2\pi\sqrt z$.
1612: We have $\hat G=G^{\rho_f}$.
1613: 
1614: By Corollary \ref{co:34-1}, $$\mathrm{Cod}(f)\approx\{2\}\cup
1615: \{[r]_{N^\rho}\in\mathrm{Rad}_{N^\rho}(f,\bar x)\,|\,\pi_r(\Sigma)
1616: \ne\{2\}\}/\hat G,$$
1617: where $2:\Sigma\to J$ is the coding map with image $\{2\}$.
1618: Let us identify $\mathrm{Rad}_{N^\rho}(f,\bar x)
1619: =(\phi^{-1}f^{-1}(\bar x))^2$ with $(\{\pm1\}\times \mathbb Z)^2$ 
1620: by $\pm1/4+n\leftrightarrow (\pm1,n)$.
1621: The action of $\mathrm{Iso}(2\mathbb Z)$ on 
1622: $(\{\pm1\}\times\mathbb Z)^2$ 
1623: is given by 
1624: \begin{multline*}
1625: (x\mapsto ax+2n)\cdot((\epsilon_1,n_1),(\epsilon_2.n_2))\\
1626: =((\epsilon_1,an_1+(2-\epsilon_1)n),
1627: (\epsilon_2,an_2+(2-\epsilon_2)n) .
1628: \end{multline*}
1629: It is easily seen that $\pi_r(\Sigma)=\{2\}$ if and only if
1630: $[r]_{N^\rho}=(b_1,b_2)$, $b_1,b_2\in\{(1,n),(-1,3n)\}$ for some $n\in
1631: \mathbb Z$.
1632: Thus
1633: \begin{multline*}
1634: \mathrm{Cod}(f)\approx\bigl\{[(1,0),(1,0)]\bigr\}
1635: \cup
1636: \big\{[(\epsilon_1,0),(\epsilon_2,n)]\,:\,\epsilon_i\in\{\pm1\},
1637: n\in\mathbb N\}\\
1638: \cup\{[(-1,1),(\epsilon,m)]\,:\,\epsilon\in\{\pm1\},m\in\mathbb Z\big\}.
1639: \end{multline*}
1640: 
1641: 
1642: In general, consider $f_d(z)=2T_d(z/2)$, where 
1643: $T_d$ is the Chebyshev polynomial of degree $d$ (i.e. $f_d\circ\phi(z)
1644: =\phi(\pm dz+2n)$).
1645: Then $f_2(z)=z^2-2$.
1646: For the basepoint $\bar x=0$, $\phi^{-1}f_d^{-1}(\bar x)=\{(\pm1/2+2n)/d
1647: \,|\,n\in\mathbb Z\}$.
1648: Similarly to the above, 
1649: $$\mathrm{Cod}(f_d)\approx\{2\}\,(\cup\{-2\})\cup
1650: \{[r]_{N^\rho}\in\mathrm{Rad}_{N^\rho}(f_d,\bar x)\,|\,\pi_r(\Sigma)
1651: \ne\{2\},\{-2\}\}/\hat G,$$ 
1652: where the term $\{-2\}$ appears if $d$ is odd.
1653: Letting $(\pm1/2+2n)/d$ correspond to $(\pm1,n)\in\{\pm1\}
1654: \times\mathbb Z$, the action of $\hat G=\mathrm{Iso}(2\mathbb Z)$ 
1655: on $\mathrm{Rad}_{N^\rho}(f,\bar x)\approx(\{\pm1\}\times\mathbb Z)^d$ 
1656: is given by 
1657: $$(x\mapsto ax+2n)\cdot((\epsilon_i,n_i))=((\epsilon_i,an_i
1658: +(d-\epsilon_i)n)).$$
1659: 
1660: \item
1661: $f(z)=-(z-1)^2/4z$.
1662: The group $G^{\rho_f}$ is equal to $$\mathrm{Iso}^+(2\Gamma)=
1663: \{z\mapsto a z+2b\,|\,a=\pm1,\pm i,\ b\in\Gamma\}.$$
1664: Similarly to the above, if $\pi_r(\Sigma)\ne \{\infty\}$,  
1665: $N_r=N^{\rho_f}$; otherwise $N_r/N^{\rho_f}
1666: =\langle z\mapsto iz+2b\rangle$ for some $b\in\Gamma'$.
1667: Hence $\hat G=G^{\rho_f}$.
1668: We have $$\mathrm{Cod}(f)\approx\{\infty\}\cup
1669: \{[r]_{N^\rho}\in\mathrm{Rad}_{\hat N}(f,\bar x)\,|\,\pi_r(\Sigma)
1670: \ne\{\infty\}\}/\hat G,$$
1671: where $\infty$ denotes the coding map with image $\{\infty\}$.
1672: Letting $\alpha/2+(1+i)\beta\ (\alpha\in A=\{\pm1,\pm i\},
1673: \beta\in\Gamma)$ correspond to $(\alpha,\beta)\in A\times \Gamma$, 
1674: we have the action of $\mathrm{Iso}^+(2\Gamma)$  on $(A\times 
1675: \Gamma)^2$ by 
1676: \begin{multline*}
1677: (z\mapsto a z+2b)\cdot((\alpha_1,\beta_1),
1678: (\alpha_2,\beta_2))\\
1679: =((\alpha_1,a\beta_1+(2-(1-i)\alpha_1) b),
1680: (\alpha_2,a\beta_2+(2-(1-i)\alpha_2)b)).
1681: \end{multline*}
1682: Thus 
1683: \begin{multline*}
1684:  \mathrm{Cod}(f)=\{[(1,0),(1,0)]\}\cup\{[(\alpha_1,0),(\alpha_2,\beta)]
1685: \,:\,\alpha_i\in A,\beta\in\Gamma_+\}\\
1686: \cup\{[(\alpha_1,1+i),(\alpha_2,\beta)]\,:\,\alpha_1\in\{-1,-i\},
1687: \alpha_2\in A,\beta\in\Gamma\},
1688: \end{multline*}
1689: where $\Gamma_+=\{n+mi\in\Gamma\,|\,n>0,m\ge 0\}$.
1690: \end{enumerate}
1691: \end{exa}
1692: 
1693: Let $\mathcal A(f)=\{R:\hat{\mathbb C}\to\hat{\mathbb C}
1694: \,|\,R\text{ is rational, }\deg R\ge1, R\circ f=f\circ R\}$.
1695: Then $\mathcal A(f)$ forms a monoid under composition of maps.
1696: 
1697: \begin{pro}\label{pro:32}
1698: The monoid  $\mathcal A(f)$ acts on $\mathrm{Cod}(f)=\{\pi_r\}$ 
1699: by $\pi\mapsto R\circ\pi$.
1700: Moreover, $\mathrm{mul}(R\circ \pi)=\deg R\cdot\mathrm{mul}(\pi)$ 
1701: for $\pi\in\mathrm{Cod}(f)$.
1702: \end{pro}
1703: 
1704: \begin{proof}
1705: For $R\in\mathcal A(f)$, $R(J)=J=R^{-1}(J)$.
1706: Indeed, if $x$ belongs to the Fatou set of $f$, then so does $R(x)$ 
1707: (see for example \cite{Be91}, Theorem 4.2.9).
1708: Thus $J_f=J_R$ if $\deg R\ge 2$.
1709: If $\deg R=1$, we have $J\subset R(J)$, so $R^{-1}(J)\subset J$.
1710: On the other hand, $R(J)\subset J\subset R^{-1}(J)$ since $R^{-1}\in
1711: \mathcal A(f)$.
1712: 
1713: Since $R:J\to J$ is $(\deg R)$-to-one except for finite points, 
1714: $\mathrm{mul}(R\circ \pi)
1715: =\deg R\cdot\mathrm{mul}(\pi)$.  
1716: For a radial $r$, we can assume that 
1717: $r$ does not intersect $R^{-1}(P_f)$. 
1718: It is easily seen that $R\circ\pi_r=\pi_{R\circ r}$.
1719: Thus $R\circ\pi_r\in\mathrm{Cod}(f)$.
1720: \end{proof}
1721: 
1722: \begin{pro}\label{pro:40}
1723: Every  $R\in \mathcal A(f)$ of degree $D=\deg R$
1724: bigger than one is a subhyperbolic rational map with $P_f=P_R$ 
1725: and $\rho_f=\rho_R$.
1726: Moreover, exceptional points of $f$ coincide with those of $R$.
1727: \end{pro}
1728: 
1729: \begin{proof}
1730: Recall that  $p$ is called an exceptional point of $f$ if 
1731:  $\#\bigcup_{k=0}^\infty f^{-k}(p)<\infty$.
1732: In fact, then $p$ is a critical point of degree $d=\deg f$, $f^2(p)
1733: =p$, $\bigcup_{k=0}^\infty f^{-k}(p)=\{p,f(p)\}$, and $f(p)$ is also 
1734: exceptional.
1735: 
1736: First we show that if $x=R^k(y)$ and $y$ is not an exceptional point of 
1737: $f$, then there exist $m>0$ and $z\in f^{-m}(x)$
1738: such that $\deg_y(R^k)$ is a divisor of $\deg_{z}(f^m)$. 
1739: Take $m>0$ so that $\#f^{-m}(y)>2D^k-2$.
1740: Since $\#C_{R^k}\le 2D^k-2$, there exists $w\in f^{-m}(y)$ such that 
1741: $\deg_w(R^k)=1$.
1742: Setting $z=R^k(w)$, 
1743: we have $\deg_y(R^k)\deg_w(f^m)=\deg_w(R^kf^m)=\deg_w(f^mR^k)
1744: =\deg_{z}(f^m)\deg_{w}(R^k)
1745: =\deg_{z}(f^m)$.
1746: 
1747: Second we show that if $y$ is an exceptional point of $f$, then $y$ is an 
1748: exceptional point of $R$.
1749: Suppose $y$ is an exceptional point of $f$.
1750: By commutativity, $R(y),R^2(y),\dots$ are exceptional points of $f$.
1751: Since a rational map has at most two exceptional points, $R^3(y)=R(y)$ 
1752: and $f^{-2}(R(y))=\{R(y)\}$.
1753: Assume that $R(y)$ is not an exceptional point of $R$.
1754: Then $X_k=R^{-2k}(R(y))-\{R(y)\}$ is nonempty for $k\in\mathbb N$.
1755: From $R^{2k}\circ f^2(X_k)=R(y)$, we have $f^2(X_k)\subset X_k$.
1756: Hence $X_k$ contains a periodic cycle of $f$.
1757: This contradicts the fact that
1758:  the Fatou set includes at most finite periodic cycles.
1759: 
1760: From the claims above, we deduce that $P_f=P_R$ and $\rho_f=\rho_R$. 
1761: \end{proof}
1762: 
1763: \begin{co}
1764:  If $f$ has exactly one critical point $c$ with $\#\{f^k(c)\,|\,
1765: k\in\mathbb N\}=\infty$, then $\mathcal A(f)$ is generated by $f$.
1766: \end{co}
1767: 
1768: \begin{proof}
1769: Let $R\in\mathcal A(f)$.
1770: We show that $R\circ f(c)=f^k(c)$ for some $k\in\mathbb N$.
1771: To this end, we assume that 
1772: $R\circ f(c)\notin\{f^k(c)\,|\,k\in\mathbb N\}$. 
1773: Since $f(c)\in P$, $R\circ f(c)\in P$.
1774: Thus $R\circ f(c)$ is eventually periodic under the iteration of $f$.
1775: Therefore there exist $m,n\in\mathbb N$ such that 
1776: $f^m\circ R\circ f(c)=f^{m+kn}\circ R\circ f(c)=R\circ 
1777: f^{m+kn+1}(c)$ for every $k\in\mathbb N$.
1778: This implies 
1779: a contradiction that $\#R^{-1}(f^m\circ R\circ f(c))=\infty$. 
1780: 
1781: From $R\circ f(c)=f^k(c)$, we have $R(f^n(c))=f^{n+k-1}(c)$.
1782: Hence $R=f^{k-1}$ by the identity theorem.
1783: \end{proof}
1784: 
1785: \begin{co}
1786: Let $\rho=\rho_f$ be the canonical ramification function, and 
1787: $g:S_{N^\rho}\to S_{N^{\rho}}$ a contraction with $f\circ \phi_{N^\rho}
1788: \circ g=\phi_{N^\rho}$.
1789: Then $\mathcal A(f)$ is the set of rational maps $R:\hat{\mathbb C}
1790: \to\hat{\mathbb C}$ such that there exists 
1791: a branched covering $g':S_{N^\rho}\to S_{N^\rho}$ with 
1792: $R\circ \phi_{N^\rho}\circ g'=\phi_{N^\rho}$ and 
1793: $g\circ g'=g'\circ g\circ t$ for some deck transformation $t$.
1794: \end{co}
1795: 
1796: \begin{proof}
1797: If $R\in\mathcal A(f)$, then  there exists 
1798: a branched covering $g':S_{N^\rho}\to S_{N^\rho}$ with 
1799: $R\circ \phi_{N^\rho}\circ g'=\phi_{N^\rho}$ by Proposition \ref{pro:40}.
1800: From $f\circ R\circ \phi_{N^\rho}\circ g'\circ g=\phi_{N^\rho}$ 
1801: and $f\circ R\circ \phi_{N^\rho}\circ g\circ g'
1802: =R\circ f\circ \phi_{N^\rho}\circ g\circ g'=\phi_{N^\rho}$, it follows 
1803:  that $g\circ g'=g'\circ g\circ t$ for some $t$ by Proposition 
1804: \ref{pro:1}.  
1805: 
1806: Conversely, suppose there exists 
1807: a branched covering $g':S_{N^\rho}\to S_{N^\rho}$ with 
1808: $R\circ \phi_{N^\rho}\circ g'=\phi_{N^\rho}$ and 
1809: $g\circ g'=g'\circ g\circ t$ for some deck transformation $t$.
1810: Then $R\circ f\circ \phi_{N^\rho}\circ g\circ g'=\phi_{N^\rho}$
1811: and $f\circ R\circ \phi_{N^\rho}\circ g'\circ g\circ t=\phi_{N^\rho}$.
1812: Hence
1813: \begin{eqnarray*}{}
1814: R\circ f(x)&=&R\circ f\circ\phi_{N^\rho}\circ g\circ g'
1815: ((\phi_{N^\rho}\circ g\circ g')^{-1}(x))\\
1816: &=&\phi_{N^\rho}((\phi_{N^\rho}\circ g\circ g')^{-1}(x))
1817: \ =\ \phi_{N^\rho}((\phi_{N^\rho}\circ g'\circ g'\circ t)^{-1}(x))\\
1818: &=&f\circ R\circ\phi_{N^\rho}\circ g\circ g'\circ t
1819: ((\phi_{N^\rho}\circ g'\circ g\circ t)^{-1}(x))
1820: \ =\ f\circ R(x)
1821: \end{eqnarray*}
1822: for any $x\in S$.
1823: \end{proof}
1824: 
1825: \begin{defi}
1826:  A coding map $\pi\in \mathrm{Cod}(f)$ is said to be {\em prime} 
1827: if there are no $R\in\mathcal A(f)$ and no $\pi'\in\mathrm{Cod}(f)$ 
1828: such that $\pi=R\circ \pi'$ and $\deg R\ge 2$.
1829: A radial $r$ is said to be prime if $\pi_r$ is prime.
1830: 
1831: 
1832: We write $$
1833: \begin{array}{l}
1834: \mathrm{Cod}'(f)=\{\pi\in\mathrm{Cod}(f)\,|\,\text{$\pi$ is prime}\}
1835: \approx\mathrm{Cod}(f)/\mathcal A(f)\\
1836: \mathcal M=\{\mathrm{mul}(\pi)\,|\,\pi\in\mathrm{Cod}(f)\},\ 
1837: \mathcal M'=\{\mathrm{mul}(\pi)\,|\,\pi\in\mathrm{Cod}'(f)\}
1838: \end{array}$$
1839: \end{defi}
1840: 
1841: \begin{co}
1842: {\rm(1)} $0\in\mathcal M'$,
1843: {\rm (2)} $\mathcal M'-\{0\}\ne\emptyset$,
1844: {\rm(3)} $\mathcal M=\{\deg R\cdot n\,|\,R\in\mathcal A(f),n\in\mathcal 
1845: M'\}$. 
1846: In particular, $\mathcal M\supset
1847: \{d^k\cdot n\,|\,n\in\mathcal M',k=0,1,2,\dots\}$ since $f^k\in\mathcal 
1848: A(f)$.  
1849: \end{co}
1850: 
1851: \begin{exa}
1852:  We calculate $\mathcal M'$, $\mathcal M$ and $\mathrm{Cod}'(f)$ 
1853: for a couple of examples in Section \ref{sec:3}.
1854: \begin{enumerate}
1855:  \item  $f(z)=z^d$.
1856: The monoid 
1857: $\mathcal A(f)=\{e^{2\pi im/d}z^k\,|\,k\in\mathbb Z,m=0,1,\dots,d-1
1858: \}$ is identified with 
1859: $\{(m,k)\}$ (the product is given by $(m,k)(m',k')=(m+km',kk')$).
1860: The action of $\mathcal A(f)$ on $\mathrm{Cod}(f)=\mathbb Z^d/\mathbb 
1861: Z$ is given by 
1862: $$(m,k)\cdot[n_1,n_2,\dots,n_d]=[kn_1+m,kn_2+m,\dots,kn_d+m].$$
1863: Hence $$\mathrm{Cod}(f)/\mathcal A(f)\approx\{[0,n_2,\dots,n_d]\,|\,
1864: n_2,\dots,n_d\text{ are mutually prime}\}.$$
1865: Since $1\in\mathcal M'$, we have $\mathcal M=\{0\}\cup\mathbb N$.
1866: 
1867: In the case $d=2$, 
1868: $\mathrm{Cod}(f)$ is identified with $\mathbb Z$ by $n_2-n_1
1869: \leftrightarrow [n_1,n_2]$, and $\mathrm{mul}([n_1,n_2])=n_2-n_1$. 
1870: We can see that $\mathrm{Cod}'(f)$ is identified with 
1871: $\{[0,0],[0,1],[0,-1]\}$.
1872: Thus $\mathcal M'=\{0,1\}$.
1873: 
1874: In the case $d=3$, $\mathrm{Cod}'(f)\approx
1875: \{[0,n_2,n_3]\,|\,\text{$n_2$, 
1876: $n_3$ are mutually prime}\}$ is infinite, 
1877: and $\mathcal M'=\{0,1\}$.
1878: 
1879: In the case $d=4$, $\mathcal M'\supset\{0,1,2\}$ 
1880: (see \cite{LaWa96}, Example 3.1).
1881: 
1882: \item $f(z)=z^2-2$.
1883: Then $\mathrm{Cod}(f)\approx(\{+,-\}\times \mathbb Z)^2/
1884: \mathrm{Iso}(\mathbb Z)$ and $\mathcal A(f)=\{f_d\,|\,d=1,2,\dots\}
1885: \approx\mathbb N$, where $f_d(z)=2T_d(z/2)$.
1886: The action of $\mathcal A(f)$ on $\mathrm{Cod}(f)$ is given by 
1887: $n\cdot[(\epsilon_1,n_1),(\epsilon_2,n_2)]=[(\epsilon_1,nn_1),
1888: (\epsilon_2,nn_2)]$.
1889: Hence 
1890: \begin{multline*}
1891: \mathrm{Cod}'(f)\approx\big\{[(+,0),(+,0)],[(-,1),(-,1)],
1892: [(-,1),(+,0)]\big\}\\
1893: \cup\big\{[(\epsilon_1,0),(\epsilon_2,1)]\,:\,\epsilon_i\in\{+,-\}
1894: \big\}
1895: \cup\big\{[(-,1),(-,\pm3^n+1)]\,:\,n\in\mathbb N\cup\{0\}\big\}.
1896: \end{multline*}
1897: Thus $\mathcal M'=\{0,1,2\cdot3^n\,|\,n=0,1,\dots\}$ 
1898: and $\mathcal M=\{0\}\cup
1899: \mathbb N$.
1900: 
1901: \item $f(z)=-(z-1)^2/4z$.
1902: The monoid $\mathcal A(f)=\{R_b\,|\,b\in\Gamma\}\approx\Gamma$, 
1903: where $R_b$ is the rational map with 
1904: $R_b\circ\phi(z)=\phi(b z)$.
1905: The action of $\mathcal A(f)$ on $\mathrm{Cod}(f)$ is given by 
1906: $b\cdot[(\alpha_1,\beta_1),(\alpha_2,\beta_2)]=[(\alpha_1,b\beta_1),
1907: (\alpha_2,b\beta_2)]$.
1908: Hence 
1909: \begin{multline*}
1910: \mathrm{Cod}'(f)\approx\big\{[(1,0),(1,0)]\big\}
1911: \cup\big\{[(-1,1),(-i,1+i)],[(-i,1),(-1,1+i)]\big\}\\
1912: \cup\big\{[(\alpha_1,0),(\alpha_2,1)]\,:\,(\alpha_1,\alpha_2)\in A^2
1913: -\{(-1,-i),(-i,-1)\}\big\}\\
1914: \cup\big\{[(\alpha_1,1),(\alpha_2,0)]\,:\,(\alpha_1,\alpha_2)
1915: \in\{-1,-i\}\times A-\{(-1,-i),(-i,-1)\}\big\}\\
1916: \cup\big\{[(\alpha,1),(\alpha,1)]\,:\,\alpha\in\{-1,-i\}\big\}\\
1917: \cup\big\{[(-1,1),(-1,(1-2i)^n\alpha+1)]\,:\,\alpha\in A,
1918: n\in\mathbb N\cup\{0\}\big\}\\
1919: \cup\big\{[(-i,1),(-i,(1+2i)^n\alpha+1)]\,:\,\alpha\in A,
1920: n\in\mathbb N\cup\{0\}\big\}.
1921: \end{multline*}
1922: Thus $\mathcal M'=\{0,1,4\cdot5^n\,|\,n=0,1,\dots\}$ 
1923: and $\mathcal M=\{|z|^2\,:\,z\in\Gamma\}$.
1924: 
1925: 
1926: \item $f(z)=z^2-3$.
1927: Then $\mathcal A(f)$ is generated by $f$.
1928: In this case, the complete solution has not been obtained.
1929: We will show that $1,2\in \mathcal M'$ in Example \ref{exa:35} 
1930: (we conjecture $\mathcal M'=\{1,2\}$).
1931: Consequently,  $0,2^k\in \mathcal M,k\in \mathbb N$.
1932: \end{enumerate}
1933:  \end{exa}
1934: 
1935: We pose the following problem:
1936: 
1937: \medskip
1938: 
1939: \noindent{\bf Problem.}
1940: For a given $f$, determine $\mathcal M'$ and $\mathcal M$.
1941: 
1942: \medskip
1943: 
1944: It is easily seen that $\mathcal M=\{0\}\cup\mathbb N$ 
1945: if $f$ is postcritically finite polynomial map.
1946: In general case, however, it is unknown whether $1\in\mathcal M$ or not.
1947: 
1948: 
1949: \section{Equivalence relations on the word space}\label{sec:5}
1950: 
1951:  Let $r=(l_i)$ and $r'=(l_i')$ be radials with basepoints $\bar x$ and 
1952: $\bar x'$ respectively, and $N\subset G$ a subgroup.
1953: For $r'$, the notation $x_w'$, $l'_w$, $l'_\omega$ and $F'_x(\cdot)$ 
1954: are defined in a trivial way.
1955:  Let $M$ be a real number bigger than $\sup_{\omega\in\Sigma}|l_\omega| 
1956: +\sup_{\omega\in\Sigma}|l'_\omega|$, and let 
1957: $$Y'=\{\gamma\in Y\,:\,|\gamma|<M\},\quad Y_N'=Y'/\sim_N,$$
1958: where $\gamma\sim_N \gamma'$ if $[\gamma{\gamma'}^{-1}]\in  N$.
1959: It is clear that $Y_N'$ is a finite set if $N^{\rho_f}\subset N$.
1960: The equivalence class of $\gamma$ is denoted by $[\gamma]_{N}.$
1961: 
1962: \begin{pro}\label{pro:16}
1963: For $\omega=\omega_1\omega_2\cdots,\omega'=\omega_1\omega_2'\cdots
1964: \in\Sigma$, $\pi_r(\omega)=\pi_{r'}(\omega')$ 
1965: if and only if there exist curves $\gamma_0,\gamma_1,\ldots \in Y'$ 
1966: such that 
1967: \begin{equation}
1968: [l_{\omega_k}F_{x_{\omega_k}}(\gamma_k)
1969: {l'_{\omega_k'}}^{-1}\gamma_{k-1}^{-1}]\text{ is trivial}\label{eq:4}
1970: \end{equation}
1971:  for $k=1,2,\dots$.
1972: Moreover, {\rm (\ref{eq:4})} can be replaced with 
1973: \begin{equation}
1974: [l_{\omega_k}F_{x_{\omega_k}}(\gamma_k)
1975: {l'_{\omega_k'}}^{-1}\gamma_{k-1}^{-1}]\in{ N_r}.\label{eq:5}
1976: \end{equation}
1977: \end{pro}
1978: 
1979: \begin{proof}
1980:  Suppose $p=\pi_r(\omega)=\pi_{r'}(\omega')$.
1981: We take $\gamma_k=l_{\sigma^k\omega}{l'_{\sigma^k\omega'}}^{-1}$.
1982: Then $F_{x_{\omega_k}}(\gamma_k)=F_{x_{\omega_k}}(l_{\sigma^k(\omega)})
1983: {{F'}_{x'_{\sigma^k\omega'}({l'}_{\sigma^k\omega})}}^{-1}$.
1984: Thus if $p\notin P$, then $\gamma_0,\gamma_1,\dots$ satisfy the condition 
1985: above.
1986: If $p\in P$, then modify each $\gamma_k$ in a small neighborhood of
1987:  $f^k(p)$ so that (\ref{eq:4}) holds.
1988: 
1989: Conversely, suppose there exist curves $\gamma_0,\gamma_1,\dots$ 
1990: satisfying the condition (\ref{eq:5}).
1991: We write $(\omega)_k=\omega_1\omega_2\cdots\omega_k$.
1992: Taking the product of 
1993: $$\begin{array}{l}
1994: \ [l_{(\omega)_k}F_{x_{(\omega)_k}}(\gamma_k){l'_{(\omega')_k}}^{-1}
1995: l'_{(\omega')_{k-1}}F_{x_{(\omega)_{k-1}}}(\gamma_{k-1})^{-1}
1996: l_{(\omega)_{k-1}}^{-1}] \\
1997: \qquad\qquad=[l_{(\omega)_{k-1}}F_{x_{(\omega)_{k-1}}}
1998: (l_{\omega_k}F_{x_{\omega_k}}(\gamma_k)
1999: {l'_{\omega_k'}}^{-1}\gamma_{k-1}^{-1})
2000: l_{(\omega)_{k-1}}^{-1}]\in N_r
2001: \end{array}$$
2002: from $k=n$ to $1$,  we have 
2003: $[l_{(\omega)_n}F_{x_{(\omega)_n}}(\gamma_n){l'_{(\omega)_n}}^{-1}
2004: \gamma_0]\in{ N_r}.$
2005: Thus the distance between $x_{(\omega)_n}$ and $x'_{(\omega')_n}$ 
2006: is less than $|F_{x_{(\omega)_n}}(\gamma_n)|\le c^{-n}M$.
2007: \end{proof}
2008: 
2009: \begin{thm}\label{thm:28}
2010: There exists a weighted directed graph $(V,E,\alpha)$:
2011: \begin{itemize}
2012: \item The vertex set $V$ is finite.
2013: \item The edge set $E$ is finite. 
2014: Each edge $e\in E$ has its initial vertex $e^-\in V$ and its terminal 
2015: vertex $e^+\in V$.
2016: (We do not assume that $e_0^-=e_1^-$ and $e_0^+=e_1^+$ imply $e_0=e_1$.)
2017: \item The weight function $\alpha:E\to\{1,2,\dots,d\}^2$.  
2018: \end{itemize}
2019: such that for $\omega=\omega_1\omega_2\cdots,\omega'=\omega_1'\omega_2'
2020: \cdots\in\Sigma$, 
2021: $$\pi_r(\omega)=\pi_{r'}(\omega') \iff
2022: \begin{array}{l}
2023: \text{there exist $e_1,e_2,\cdots\in E$ such that } \\
2024: \text{$e_i^+=e_{i+1}^-$ and $\alpha(e_i)=(\omega_i,\omega_i') $.}
2025: \end{array}$$
2026: \end{thm}
2027: 
2028: \begin{proof}
2029: Let $V$ be the maximal subset of $Y_{N_r}'$ such that for any 
2030: $[\gamma]_{N_r}\in Y_{N_r}'$ 
2031: there exists $[\gamma']_{N_r}\in V$ with $[\gamma]_{ N_r}
2032: =[l_iF_{x_i}(\gamma'){l_j'}^{-1}]_{N_r}$ 
2033: for some $i,j\in\{1,2,\dots,d\}$.
2034: Set 
2035: $$E=\{([\gamma]_{N_r},[\gamma']_{N_r},i,j)\in V^2\times\{1,2,\dots,d\}^2 
2036: \,|\,[\gamma]_{N_r}=[l_iF_{x_i}(\gamma'){l_j'}^{-1}]_{N_r}\}.$$
2037: For $e=([\gamma]_{ N_r},[\gamma']_{ N_r},i,j)\in E$, 
2038: define $e^-=[\gamma]_{N_r}$, 
2039: $e^+=[\gamma']_{N_r}$ and $\alpha(e)=(i,j)$.
2040: 
2041: Suppose there exist $e_1,e_2,\cdots\in E$ such that 
2042: $e_i^+=e_{i+1}^-$ and $\alpha(e_i)=(\omega_i,\omega_i')$.
2043: Since there exist curves $\gamma_k,k=0,1,\dots$ such that 
2044: $[\gamma_k]_{N_r}
2045: =e_{k+1}^-$, we have $\pi_r(\omega)=\pi_{r'}(\omega')$ by proposition 
2046: \ref{pro:16}.
2047: Conversely, suppose $\pi_r(\omega)=\pi_{r'}(\omega')$.
2048: By Proposition \ref{pro:16}, there exists $[\gamma_0]_{N_r},
2049: [\gamma_1]_{N_r},\dots\in Y'$ such that $[l_{\omega_k}
2050: F_{x_{\omega_k}}(\gamma_k)
2051: {l'_{\omega_k'}}^{-1}]_{N_r}=[\gamma_{k-1}]_{N_r}$ 
2052: for $k=1,2,\dots$.
2053: Then $[\gamma_k]_{N_r}\in V$, and so $e_k=([\gamma_{k-1}]_{N_r},
2054: [\gamma_k]_{N_r},\omega_k,\omega_k'),k=1,2,\dots$ 
2055: satisfy the condition.
2056: \end{proof}
2057: 
2058: \begin{co}\label{co:34}
2059:  Let $V$ be the vertex set constructed in Theorem \ref{thm:28}.
2060: If $\pi(\Sigma)=J$, then $$\#\pi^{-1}(x)\le 
2061: \max_w\#\{w'\,|\,x_w=x_{w'}\} \cdot \max_{p\in C\cap J,k\ge 1}
2062: \deg_p f^k\cdot\sqrt{2\#V}$$
2063: for any $x\in J$.
2064: \end{co}
2065: 
2066: \begin{proof}
2067:  An immediate consequence of  (\ref{eq:l}).
2068: \end{proof}
2069: 
2070: Is is easily seen that if $\pi(\Sigma)=J$, then 
2071: the Julia set $J$ is topologically identified 
2072: with the quotient space $\Sigma/\sim_r$, where the equivalence relation 
2073: $\sim_r$ is defined by $\omega\sim_r\omega'\iff \pi_r(\omega)
2074: =\pi_r(\omega')$.
2075: Considering $r=r'$, we obtain from Theorem \ref{thm:28} 
2076:  an algorithm to calculate $\sim_r$ 
2077: provided $Y'$ is determined. 
2078: 
2079: \begin{exa}\label{exa:35}
2080:  Let $f(z)=z^2-3$.
2081: In order to obtain $\sim_r$ for the radials $r=r_1,r_2,r_3$ in Section 
2082: \ref{sec:3.1}, we use Theorem \ref{thm:28}.
2083: Let us set generators $[B_1],[B_2],\dots$ of $G$ as shown in 
2084: Figure \ref{fig:fig5}.
2085: Then $[B_k^2],[(B_1B_2)^4]\in\hat N,k=1,2,\dots$
2086: Take a simply connected open domains $U_1\subset U_2$ 
2087: such that $B_1\subset U_1,B_2\subset U_2,
2088: [-3,3]\subset U_1,[-3,6]\subset U_2,
2089: [6,\infty)\subset\mathbb C-U_1,[33,\infty)\subset \mathbb C-U_2$, 
2090: and $f^{-1}(\overline {U_i})\subset U_i$.
2091: For the radial $r=r_j$ above, we take $r$ so that the image of $r$ 
2092: is included in $U_i$ ($i=1$ if $j=1,2$, $i=2$ if $j=3$) 
2093: without loss of generality.
2094: For any $k$, there exists $n$ such that 
2095: $F_x(B_k)\subset U_2$ for each $x\in f^{-n}(\bar x)$.
2096: Therefore, for $r=r_1,r_2$, 
2097: the set $V$ in Theorem \ref{thm:28} is included 
2098: $\{1_{\hat N},[B_1]_{\hat N}\}$, 
2099: where $1_{\hat N}=\hat N$ is the unit element.
2100: If $r=r_3$, $V$ is included in 
2101: \begin{multline*}
2102: \{1_{\hat N},[B_1]_{\hat N},[B_2]_{\hat N},[B_1B_2]_{\hat N},
2103: [B_2B_1]_{\hat N},[B_1B_2B_1]_{\hat N},[B_2B_1B_2]_{\hat N},\\
2104: [(B_1B_2)^2]_{\hat N}[(B_2B_1)^2]_{\hat N}\},
2105: \end{multline*}
2106: the subgroup generated by $[B_1]_{\hat N}$ and $[B_2]_{\hat N}$.
2107: \begin{enumerate}
2108:  \item $r=r_1$.
2109: We have 
2110: $$\begin{array}{lcll}
2111: e&\leftarrow&(11)\ e,&(22)\ e\\
2112: B_1&\leftarrow& (12)\ e,&(21)\ e
2113: \end{array}$$
2114:  where $e$ denotes the trivial loop.
2115: (For example, ``$B_1\leftarrow(12)\ e,\ (21)\ e$'' indicates that 
2116: $l_1F_{x_1}(B_1)l_2^{-1}$ and $l_2F_{x_2}(B_1)l_1^{-1}$ are  
2117:  trivial.)
2118: Thus $$V=\{1_{\hat N}\},\ E=\{(1_{\hat N},1_{\hat N},1,1),
2119: (1_{\hat N},1_{\hat N},2,2)\}.$$
2120: It follows from this that the equivalence relation $\sim_r$ is trivial,
2121: that is, $\omega\sim_r\omega'$ if and only if $\omega=\omega'$.
2122: Consequently, $\pi_r$ is bijective.
2123: \item $r=r_2$.
2124: We have 
2125: $$
2126: \begin{array}{lcll}
2127: e&\leftarrow&(11)\ e,&(22)\ e\\
2128: B_1&\leftarrow& (12)\ B_1^{-1},&(21)\ B_1\\
2129: \end{array}$$
2130: (``$B_1\leftarrow(11)\ B_1^{-1},\ (22)\ B_1$'' indicates that 
2131: $l_1F_{x_1}(B_1)l_2^{-1}$ is homotopic to $B_1^{-1}$ and 
2132: $l_2F_{x_2}(B_1)l_1^{-1}$ is homotopic to $B_1$).
2133: Thus $$\begin{array}{l}V=\{1_{\hat N},[B_1]_{\hat N}\},\\
2134:  E=\{(1_{\hat N},1_{\hat N},1,1),(1_{\hat N},1_{\hat N},2,2),
2135: ([B_1]_{\hat N},[B_1]_{\hat N},1,2),
2136: ([B_1]_{\hat N},[B_1]_{\hat N},2,1)\}.
2137: \end{array}$$
2138: It follows from this that 
2139: $$\omega=\omega_1\omega_2\cdots\sim_r \omega'=\omega_1'\omega_2'\cdots
2140:  \iff \omega=\omega'\text{ or }\omega_k\ne\omega_k',k=1,2,\dots$$
2141: (for example, $111\cdots\sim_r222\cdots$ and 
2142:  $1212\cdots\sim_r2121\cdots$). 
2143: Consequently, $\pi_r$ is exactly two-to-one.
2144: \item $r=r_3$.
2145: We have 
2146: $$
2147: \begin{array}{lcll}
2148: e&\leftarrow&(11)\ e,&(22)\ e\\
2149: B_1&\leftarrow& (12)\ B_2,&(21)\ B_2^{-1}\\
2150: B_2&\leftarrow& (11)\ e,&(22)\ B_2^{-1}B_1B_2\\
2151: B_1B_2&\leftarrow& (12)\ B_1B_2,&(21)\ B_2^{-1}\\
2152: B_2B_1&\leftarrow& (12)\ B_2,&(21)\ B_2^{-1}B_1\\
2153: B_2B_1B_2&\leftarrow& (12)\ B_1B_2,&(21)\ B_2^{-1}B_1
2154: \end{array}$$
2155: etc.
2156: Thus 
2157: $$\begin{array}{ccl}
2158:  V&=&\{1_{\hat N}, [B_2]_{\hat N},[B_1B_2]_{\hat N},[B_2B_1]_{\hat N},
2159: [B_2B_1B_2]_{\hat N}\},\\
2160:  E&=&\{(1_{\hat N},1_{\hat N},1,1),(1_{\hat N},1_{\hat N},2,2),\\
2161: &&\qquad(1_{\hat N},[B_2]_{\hat N},1,1),
2162: ([B_2]_{\hat N},[B_2B_1]_{\hat N},1,2),\\
2163: &&\qquad ([B_2]_{\hat N},[B_1B_2]_{\hat N},2,1),
2164: ([B_2B_1]_{\hat N},[B_2B_1]_{\hat N},2,1),\\
2165: &&\qquad([B_1B_2]_{\hat N},[B_1B_2]_{\hat N},1,2),
2166:  ([B_2B_1]_{\hat N},[B_2B_1B_2]_{\hat N},2,1),\\
2167: &&\hspace{1.5cm}([B_1B_2]_{\hat N},[B_2B_1B_2]_{\hat N},1,2),
2168: ([B_2B_1B_2]_{\hat N},[B_2]_{\hat N},2,2)\}.
2169: \end{array}$$
2170: By Corollary \ref{co:34}, $\pi_r$ is at most three-to-one.
2171: The graph $(V,E,\alpha)$ is diagrammatically shown as
2172: $$\begin{array}{crclc}
2173: {\scriptstyle 11,22}\ \text{\scalebox{1.5}{\rotatebox[origin=c]{270}{$\circlearrowleft$}}\ }
2174: 1&\xrightarrow{\ 11\ } \quad B_2\ &\xrightarrow{\ 12\ }&B_2B_1\
2175:  \text{\scalebox{1.5}{\rotatebox[origin=c]{90}{$\circlearrowleft$}}}
2176: \ {\scriptstyle 21}\\
2177: &{\scriptstyle 21}\downarrow\ \ \ & \ \nwarrow{\scriptstyle 22}   &
2178: \ \ \downarrow{\scriptstyle 21}      \\
2179: &{\scriptstyle 12}
2180: \ \text{\scalebox{1.5}{\rotatebox[origin=c]{270}{$\circlearrowleft$}}\ }
2181: B_1B_2& \xrightarrow[\ 12\ ]{}& B_2B_1B_2
2182: \end{array}$$
2183: We can see that if $\omega\in\Sigma$ contains a word $12121$ 
2184:  infinitely many times, then $\pi_r^{-1}\pi_r(\omega)=\{\omega\}$.
2185: Consequently, the multiplicity of $\pi_r$ is equal to one.
2186: \end{enumerate}
2187: \begin{figure}[hbtp]
2188: \centering
2189: \includegraphics[height=5cm]{fig5.eps}
2190: \caption{Generators of $G$.}
2191: \label{fig:fig5}
2192: \end{figure}
2193: 
2194: \end{exa}
2195: 
2196: 
2197: 
2198: 
2199: \section{Non-subhyperbolic case}\label{sec:nonhyp}
2200: A couple of our results are true for non-subhyperbolic rational maps.
2201: In fact, if $f$ is geometrically finite (i.e. $J\cap P$ is finite),
2202: then almost all of our results are applicable, but 
2203: it is possible that $J\not\subset S$.
2204: Thus the Julia tile $K$ might be noncompact.
2205: The details are left to the reader.
2206: 
2207: For general rational maps, we need some restriction. 
2208: Let $f:\hat{\mathbb C}\to\hat{\mathbb C}$ be a rational map.
2209: If $S'=\hat{\mathbb C}-P$ has a connected component $U$ such that 
2210: $f^{-1}(U)\subset U$, then we have 
2211: a radial $r$ in $U$,
2212: but $x_\omega$ does not converge to a point in general.
2213: We define
2214:  $$\Pi(\omega)=\bigcap_{k=1}^\infty \overline{
2215: \bigcup_{\omega'\in\Sigma}
2216: F_{x_{\omega_1\omega_2\cdots\omega_{k}}}(l_{\omega'})}$$
2217: for $\omega=\omega_1\omega_2\cdots\in\Sigma$.
2218: Then we have
2219: 
2220: \begin{pro}
2221:  {\rm(1)} $f(\Pi(\omega))=\Pi({\sigma\omega})$,
2222: {\rm (2)} $\Pi(\omega)\subset J$,  
2223: {\rm (3)} $\omega\mapsto \Pi(\omega)$ is upper semicontinuous, that is,  
2224: if a sequence $\omega^1,\omega^2,\dots$ in $\Sigma$ converges to 
2225: $\omega$, then $\bigcap_{m=1}^\infty\overline{\bigcup_{k=m}^\infty
2226: \Pi({\omega^k})}\subset \Pi(\omega)$ 
2227:  and {\rm (4)} $\Pi(\omega)$ is connected.
2228: \end{pro}
2229: 
2230: \begin{proof}
2231: (1) is immediately deduced from the definition.
2232: 
2233: (2): Suppose a point $x\in\Pi(\omega)$ belongs to the Fatou set.
2234: It is easily seen that $f^n(x)$ converges to neither a 
2235: (super)attracting cycle nor a parabolic cycle.
2236: Thus $f^n(x)\in \Pi(\sigma^n\omega)$ for some $n$ is contained in 
2237: either a Siegel disc or a Herman ring. 
2238: This contradicts the fact that  $f:f^{-1}(U)\to U$ is expanding 
2239: in the Poincar\'e metric on $U$.  
2240: 
2241: (3): Set
2242:  $\omega=\omega_1\omega_2\cdots$ and $\omega^k=\omega_1^k\omega_2^k
2243: \cdots$.
2244: For any $n$, there exists 
2245: $m_0$ such that $\omega^m\in \Sigma(\omega_1\omega_2\cdots\omega_n)$ 
2246: whenever $m\ge m_0$.
2247: Since $$\Pi(\omega^m)\subset
2248: \overline{\bigcup_{\omega'\in\Sigma}
2249: F_{x_{\omega_1^m\omega_2^m\cdots\omega_{n}^m}}(l_{\omega'})}=
2250: \overline{\bigcup_{\omega'\in\Sigma}
2251: F_{x_{\omega_1\omega_2\cdots\omega_{n}}}(l_{\omega'})}$$ for $m\ge m_0$,
2252: we have $\bigcap_{m=1}^\infty\overline{\bigcup_{k=m}^\infty
2253: \Pi({\omega^k})}\subset 
2254: \overline{\bigcup_{\omega'\in\Sigma}
2255: F_{x_{\omega_1\omega_2\cdots\omega_{n}}}(l_{\omega'})}$ for every $n$. 
2256: 
2257: (4): $\Pi(\omega)$ is connected since 
2258: $\overline{\bigcup_{\omega'\in\Sigma}
2259: F_{x_{\omega_1\omega_2\cdots\omega_{k}}}(l_{\omega'})}$ is connected.
2260: \end{proof} 
2261: 
2262: From (2) and (4) immediately, 
2263: 
2264: \begin{co}
2265: If $J$ is totally disconnected, then $\Pi(\omega)$ is a singleton for 
2266: every $\omega\in\Sigma$.
2267: \end{co}
2268: 
2269: \begin{rem}
2270: In general, by Przytycki's result \cite{Prz86},  
2271: $\Pi(\omega)$ is a singleton for every $\omega\in\Sigma$ 
2272: except for $\omega$ in a ``thin'' set.
2273: \end{rem}
2274: 
2275: \begin{thm}
2276: The multiplicity of $\Pi$ is well-defined, that is, there exists 
2277: $n_r\in\mathbb N\cup \{\infty\}$ such that 
2278: $\#\{\omega\in\Sigma\,|\,x\in\Pi(\omega)\}=n_r$ 
2279: for $\mu$-almost all $x\in J$.
2280: \end{thm}
2281: 
2282: \begin{proof}
2283: It is sufficient to show that the function 
2284: $h(x)=\#\{\omega\in\Sigma\,|\,x\in\Pi(\omega)\}$ is Borel. 
2285: Let $A_{k,\epsilon}$ be the set of $x\in J$ 
2286: such that for some $\omega^i\in K,i=1,2,\dots,k$, 
2287: $x\in \Pi(\omega^i),1\le i\le k$ and
2288: the distance between $\omega^i$ and $\omega^j$ is  equal to or bigger 
2289: than $\epsilon$ for $0\le i\ne j\le k$, where we consider an  
2290: arbitrary compatible distance function on $\Sigma$. 
2291: Then $A_{k,\epsilon}$ is closed since $\omega\mapsto\Pi(\omega)$ is 
2292: upper semicontinuous.
2293: Thus $\{h(x)\ge k\}$ is Borel.
2294: \end{proof}
2295: 
2296: 
2297: 
2298: \begin{thebibliography}{10}
2299: 
2300: \bibitem{Be91}
2301: A.~F. Beardon, \emph{Iteration of rational functions}, Graduated Text in
2302:   Mathematics, Vol.132, Springer-Verlag, 1991.
2303: 
2304: \bibitem{Bo75}
2305: R.~Bowen, \emph{Equilibrium states and the ergodic theory of {A}nosov
2306:   diffeomorphisms}, Lec. Notes in Math., vol. 470, Springer-Verlag, Berlin,
2307:   1975.
2308: 
2309: \bibitem{CoEc80}
2310: P.~Collet and J.-P. Eckmann, \emph{Iterated maps of the interval as dynamical
2311:   systems}, Birkhauser, 1980.
2312: 
2313: \bibitem{MeSt93}
2314: W.~de~Melo and S.~van Strien., \emph{One-dimensional dynamics},
2315:   Springer-Verlag, 1993.
2316: 
2317: \bibitem{Fri87}
2318: D. Fried, \emph{Finitely presented dynamical systems}, Ergod. Th. Dynam.
2319:   Sys. (1987), 489--507.
2320: 
2321: \bibitem{Gu70}
2322: J. Guckenheimer, \emph{Endomorphisms of the {R}iemann sphere}, Global
2323:   analysis, Proc. Sympos. Pure Math, vol. XIV, AMS, 1970, pp.~95--123.
2324: 
2325: \bibitem{Hu81}
2326: J.~E. Hutchinson, \emph{Fractals and self-similarity}, Indiana Univ. Math. J.
2327:   \textbf{30} (1981), 713--747.
2328: 
2329: \bibitem{Ja68}
2330: M.~V. Jakobson, \emph{Structure of polynomial mappings on a singular set},
2331:   Math. USSR Sbornik \textbf{6} (1968), 97--114.
2332: 
2333: \bibitem{Ja69}
2334: \bysame, \emph{On the problem of classification of polynomial endomorphisms of
2335:   the plane}, Math. USSR Sbornik \textbf{9} (1969), 345--364.
2336: 
2337: \bibitem{Ka03}
2338: A. Kameyama,
2339: \emph{On {J}ulia sets of postcritically finite branched coverngs 
2340: {P}art {I}-- coding of {J}ulia sets},
2341: J. Math. Soc. Japan, \textbf{55} (2003), 439--454.
2342: 
2343: \bibitem{Ken92}
2344: R.~Kenyon, \emph{Self-replicating tilings}, Symbolic dynamics and its
2345:   applications (P.~Walters, ed.), Contemp. Math., vol. 135, AMS, pp.~239--263.
2346: 
2347: \bibitem{Ken97}
2348: \bysame, \emph{Projecting the one-dimensional {S}ierpinski gasket},
2349:   Israel J. Math. \textbf{97} (1997), 221--238.
2350: 
2351: \bibitem{LaWa96}
2352: J.~C. Lagarias and Y. Wang, \emph{Integral self-affine tiles in $\mathbb
2353:   {R}^n$ {I}. {S}tandard and nonstandard digit sets}, J. London Math. Soc.
2354:   \textbf{54} (1996), 161--179.
2355: 
2356: \bibitem{LaWa96b}
2357: \bysame, \emph{Self-affine tiles in $\mathbb {R}^n$}, Adv. Math. \textbf{121}
2358:   (1996), 21--49.
2359: 
2360: \bibitem{Lj83}
2361: M.~Ju. Ljubich, \emph{Entropy properties of rational endomorphisms of the
2362:   {R}iemann sphere}, Ergod. Th. Dynam. Sys. \textbf{3} (1983), 351--385.
2363: 
2364: \bibitem{Ly86}
2365: \bysame, \emph{The dynamics of rational transforms: the topological
2366:   picture}, Russian Math. Surveys \textbf{41:4} (1986), 43--117.
2367: 
2368: \bibitem{Mi99}
2369: J.~Milnor, \emph{Dynamics in one complex variable: Introductory lectures},
2370:   Vieweg, Braunschweig/Wiesbaden, 1999.
2371: 
2372: \bibitem{MiTh88}
2373: J.~Milnor and W.~Thurston, \emph{On iterated maps of the interval}, Dynamical
2374:   Systems (J.~C. Alexander, ed.), Lect. Notes in Math., no. 1342, Springer,
2375:   1988, pp.~465--563.
2376: 
2377: \bibitem{Prz85}
2378: F. Przytycki, \emph{Hausdorff dimension of harmonic measure on the boundary
2379:   of an attractive basin for a holomorphic map}, Invent. Math. \textbf{80}
2380:   (1985), 161--179.
2381: 
2382: \bibitem{Prz86}
2383: \bysame, \emph{Riemann map and holomorphic dynamics}, Invent. Math. \textbf{85}
2384:   (1986), 439--455.
2385: 
2386: \bibitem{PrSk91}
2387: \bysame and J. Skrzypczak, \emph{Convergence and pre-images of limit
2388:   points for coding trees for iterations of holomorphic maps}, Math. Ann.
2389:   \textbf{290} (1991), 425--440.
2390: 
2391: \end{thebibliography}
2392: \end{document}
2393: