959 1996 112-126

112

On the Boundary of unbounded invariant Fatou

Components of Entire Functions
Masashi KISAKA (
)

Department of Mathematics, College of Integrated Arts and Science, University of Osaka Prefecture, Gakuen-cho 1-1, Sakai 593, Japan -mail: kisaka@mathsun.cias.osakafu-u.ac.jp

Denitions and Results

$f$
$f^{n}$

denote the n-th Let be a transcendental entire function and of are and the Julia set iterate of . Recall that the Fatou set dened as follows:
$f$
$F_{f}$
$J_{f}$

$f$

$F_{f}:=$

$z\in \mathbb{C}|\{f^{n}\}_{n=1}^{\infty}$

is a normal family in a neighborhood of },

$z$

$J_{f}:=\mathbb{C}\backslash F_{f}$

.
$\infty$

$\overline{|}r_{:=}.\mathbb{C}\cup\{\infty\}$

It is possible to consider the Julia set to be a subset of the Riemann sphere to it. This denition is by adding the point of innity mainly adopted in the case of meromorphic functions (for example, see [Ber] and also there are some researches on convergence phenomena of is compact ). In this setting, ([Ki], [Kr], Julia sets as subsets is rather easy to handle. But for a transcendental entire in and hence function the suitable phase space as a dynamical system is the complex is an essential singularity plane , not the Riemann sphere , because of and there seems to be no reasonable way to dene the value at . as a subset of rather than of and So it is more natural to regard as above and write hence we dene } when we consider to be a subset of . A connected component of is called a Fatou component. A Fatou component is called a wandering domain if $f^{m}(U)\cap f^{n}(U)=\emptyset$ for every is called with $m,$ $n\in \mathrm{N}(m\neq n)$ . If there exists an a periodic component and it is well known that there are following four possibilities:
$)$
$\mathrm{o}\mathrm{f}\overline{\mathbb{C}}$

$[\mathrm{K}\mathrm{r}\mathrm{K}]$

$J_{f}$

$\overline{\mathbb{C}}$

$J_{f}$

$\overline{\mathbb{C}}$

$\mathbb{C}$

$\infty$

$f$

$\infty$

$\overline{\mathbb{C}}$

$J_{f}$

$\mathbb{C}$

$J_{f}$

$J_{f}\cup \mathrm{t}\infty$

$J_{f}$

$\overline{\mathbb{C}}$

$U$

$F_{f}$

$n_{0}\in \mathrm{N}$

$f^{n_{0}}(U)\subseteq U,$

$U$

113

1. There exists a point $z_{0}\in U$ with $f^{n_{0}}(z_{0})--z_{0}$ and as $karrow\infty$ . The point and every point $z\in U$ satises is called $an$ is called an attracting periodt,c point and the domain . attracting basin. $*.1$
$|(f^{n_{0}})(z\mathrm{o})|<1$
$z_{0}$

$f^{n_{0}k}(z)arrow z_{0}$

$U$

$\wedge\cdot$

$\mathrm{t}$

$\sim$

with $f^{n_{0}}(z_{0})=z_{0}$ and $(f^{n_{0}})(z_{0})=$ 2. There exists a point as $karrow\infty$ . and every point $z\in U$ satises is called a parabolic periodic point and the domain is The point called a parabolic basin.
$z_{0}\in\partial U$
$e^{2\pi i\theta}(\theta\in \mathbb{Q})$

$f^{n_{0}k}(z)arrow z_{0}$

$z_{0}$

$U$

3. There exists a

$\sim \mathrm{p}\mathrm{o}\mathrm{i}\overline{\mathrm{n}}\mathrm{t}z_{0}\in U^{\vee}$

and of a unit disk. The domain

$e^{2\pi i\theta}(\theta\in \mathbb{R}\backslash \mathbb{Q})$

$f^{n_{0}}|U$

$U$

with $f^{n_{0}}(z_{0})=z_{0}$ and is conjugate to an irrational rotation : is called a Siegel disk.

$(f^{n_{0}})(z\mathrm{o})=$

4. For every $z\in U,$ Baker domain.

$f^{n_{0}k}(z)arrow\infty$

as

$karrow\infty$

. The domain

$U$

is called

$a$

component. is called is called an inva In particular, if $n_{0}=1,$ $f ^{-1}(U)\subseteq U.$ is called a preperisatises completely invariant if is odic component if $f^{m}(U)$ is a periodic component for an $m\geq 1$ . is periodic or preperiodic. It is known odic if called eventually that eventually periodic components of a transcendental entire function while a wandering domain can be are simply connected multiply-connected ([Bal], [Ba2], [Ba5]). The boundary of unbounded periodic Fatou component can be extremely complicated. For example, consider the exponential family has a unique attracting . If satises with an unbounded simply connected completely invariant xed point . Let is equal to this basin basin and the Fatou set from a unit disk , then the : be a Riemann map of and moreover the set exists for all radial limit
$U$
$7\dot{\mathrm{v}}ant$

$U$

$U$

$U$

$U$

$per\dot{\mathrm{v}}$

$U$

$([\mathrm{B}\mathrm{e}\mathrm{r}], [\mathrm{E}\mathrm{L}1])$

$E_{\lambda}(z):=\lambda e^{z}$

$\lambda$

$0< \lambda<\frac{1}{e},$

$E_{\lambda}(z)$

$p_{\lambda}$

$\Omega(p_{\lambda})$

$F_{E_{\lambda}}$

$([\mathrm{D}\mathrm{G}])$

$\varphi$

$\mathrm{D}arrow\Omega(p_{\lambda})$

$\Omega(p_{\lambda})$

$\mathrm{D}$

$\lim_{r\nearrow 1\varphi}(re^{i})\theta$

$e^{i\theta}\in\partial \mathrm{D}$

$\Theta_{\infty}:=\{e^{i\theta}|\varphi(e^{i\theta}):=\lim_{r\nearrow 1}\varphi(rei\theta)--\infty\}$

. This implies that the Riemann map is highly is dense in , is , which is equal to discontinuous and hence the boundary of is disconnected extremely complicated. Rom this fact, it folows that in , since is conformal the set
$\partial \mathrm{D}([\mathrm{D}\mathrm{G}])$

$\Omega(p_{\lambda})$

$J_{E_{\lambda}}$

$J_{E_{\lambda}}$

$\mathbb{C}$

$\varphi$

$\varphi(\{re^{i\theta_{1}}|0\leq r<1\}\cup\{re^{i\theta_{2}}|0\leq r<1\})\subset U$

$(\theta_{1}, \theta_{2}\in\Theta_{\infty}, \theta_{1}\neq\theta_{2})$

114

is a Jordan arc in and this separates into two disjoint relatively open subsets. Taking these facts into account, we shall investigate the set for a genetal unbounded periodic component and also consider the following problem
$\mathbb{C}$

$J_{E_{\lambda}}$

$\Theta_{\infty}$

$U$

Problem : When is the Julia set of a transcendental entire function connected or disconnected as a subset of ?
$\mathbb{C}$

$f$

is a polynomial, the following criterion is well known. (For example, see [Bea] or [M] .

If

$f$

$)$

Proposition A Let be a polynomial of degree $d\geq 2$ . Then the Julia set is connected if and only if no nite values of tend to by the iterates of .
$f$
$J_{f}$

$c\dot{n}tiCa\iota$

$f$

$\infty$

$f$

Here, a $crit8,cal$ value is a point $p:=f(c)$ for a point with $f(c)=0$ . This is a singularity of . For polynomials we have only to consider this type of singularities but there can be another type of singularities called an asymptotic value for the transcendental case. A point is called an asymptotic value if there exists a continuous curve $L(t)(0\leq t<1)$ called an asymptotic path with
$c$

$f^{-1}$

$p$

$\lim_{tarrow 1}L(t)=\infty$

and

$\lim_{tarrow 1}f(L(t))=p$

A point is called a singular value if it is either a critical or an asymptotic value and we denote the set of all singular values as sing $(f^{-}1)$ . If is transcendental, however, the above criterion does not hold. For example, let us consider the exponential family again. If , the unique singular value $z=0$ (this is an asymptotic satises but value) is attracted to the xed point and hence does not tend to is disconnected as we mentioned above. the Julia set , the singular value $z=0$ may For other values of , for example tend to . If is a polynomial all of whose critical values tend to , then is a Cantor set and especially disconnected. But on the other hand in and hence connected. this case is equal to the entire plain in , we investigate the Before considering the connectivity of this situation compactness of in . connectivity of in makes the problem easier. Actually we can prove the following:
$p$

$f$

$E_{\lambda}(z):=\lambda e^{z}$

$\lambda$

$0< \lambda<\frac{1}{e}$

$p_{\lambda}$

$\infty$

$J_{E_{\lambda}}$

$\lambda$

$\lambda>\frac{1}{e}$

$\infty$

$f$

$\infty$

$J_{f}$

$J_{f}$

$\mathbb{C}([\mathrm{D}])$

$J_{f}$

$\mathbb{C}$

$J_{f}\cup\{\infty\}$

$\overline{\mathbb{C}}$

$\ln$

$J_{f}\cup\{\infty\}$

$\overline{\mathbb{C}}$

115

Theorem 1 Let be a transcendental entire function. Then the set in is connected if and only if has no multiply-connected wandering domains.
$f$
$J_{f}\cup\{\infty\}$
$\overline{\mathbb{C}}$

$F_{f}$

Corollary 1 Under one of the following conditions, connected. (1) $f\in B:=$ { is bounded}. (2) has an unbounded component. (3) There exists a curve $\Gamma(t)(0\leq t<1)$ with is bounded. Especially $f$ has a nite asymptotic value.
$f|\mathrm{s}\mathrm{i}\mathrm{n}\mathrm{g}(f^{-}1)$

$J_{f}\cup\{\infty\}$

in

$\overline{\mathbb{C}}$

is

$F_{f}$

$\lim_{tarrow 1}\Gamma(t)=\infty$

such that

$f|\Gamma$

in Then how about itself? The results depend on whether admits no admits an unbounded component or not. In the case when unbounded components, we obtain the following:
$J_{f}$
$\mathbb{C}$

$F_{f}$

$F_{f}$

Theorem 2 Let be a transcendental entire function. ponents of are bounded and simply connected, then
$f$
$F_{f}$

If all
$J_{f}$

the comis connected.

The following is an easy consequence from Theorem 1 and 2.

Corollary 2 Let $f$ be a transcendental entire function. If all the components of are bounded, then is connected in if and only if is connected in .
$F_{f}$
$J_{f}$
$\mathbb{C}$

$J_{f}\cup\{\infty\}$

$\overline{\mathbb{C}}$

of As we mentioned before, for the unbounded component admits the the set of all angles where the Riemann map : . radial limit is dense in and this leads to the disconnectivity of The Main result of this paper is the generalization of this fact. Under some conditions this result holds for various kinds of unbounded periodic is accessible if there exists a Fatou components. Here, a point continuous curve $L(t)(0\leq t<1)$ in with $\lim_{tarrow 1}L(t)=p$ .
$\Omega(p_{\lambda})$
$F_{E_{\lambda}}$
$\varphi$

$\mathrm{D}arrow\Omega(p_{\lambda})$

$\infty$

$\partial \mathrm{D}$

$J_{E_{\lambda}}$

$p\in\partial U$ $U$

Main Theorem Let be an unbounded : transcendental entire function $f,$ from a unit disk and
$U$
$\varphi$
$\mathrm{D}_{f}$

$pe$

$\mathrm{D}arrow U$

riodic Fatou component of a be a Riemann map of

$U$

$P_{f^{n_{0}}}:= \bigcup_{n=0}^{\infty}(fn_{0})n(\mathrm{s}\mathrm{i}\mathrm{n}\mathrm{g}((f^{n_{0}})-1))$

.
is accessible. There

We assume one of the following four conditions: and is an attracting basin of period (1)
$U$
$n_{0}$

$\infty\in\partial U$

116

$C(t)\subset U(0\leq t\leq 1)$

$U$

with $C(1)=q$ and satises (2) is a parabolic basin of and exists a nite point with $C(t)\subset U(0\leq t\leq 1)$ with $C(1)=q$ and satises $f^{m_{0}}(C)\supset C$ . (3) is a Siegel disk of period $n0$ and is accessible. (4) is a Baker domain of period $n0$ and is not univalent. There exists a nite point with and a continuous curue $C(t)\subset U(0\leq t\leq 1)$ with $C(1)=q$ and satises $f^{m_{0}}(C)\supset C$ . Then the set
$per\dot{\mathrm{v}}odn0$

exists a nite point

$q\in\partial U$

with

$q\not\in P_{f^{n_{0}}},$

$m_{0}\in \mathrm{N}$

$\infty\in\partial U$

$q\in\partial U$

$q\not\in P_{f^{n_{0_{f}}}}m_{0}\in \mathrm{N}$

and a continuous curue $f^{m_{0}}(C)\supset C$ . is accessible. There and a continuous curve

$U$

$\infty\in\partial U$

$U$

$f^{n_{0}}|U$

$q\in\partial U$

$q\not\in P_{f^{n_{0_{f}}}}m_{0}\in \mathrm{N}$

$\Theta_{\infty}:=\{e^{i\theta}|\varphi(e^{i\theta}):=\lim_{r\nearrow 1}\varphi(rei\theta)=\infty\}$

$\overline{\Theta_{\infty}}$

is dense in in the case of (1), (2) or (3). In the case . In particular, contains a certain perfect set in in all cases.
$\partial \mathrm{D}$ $\partial \mathrm{D}$

of (4),
$J_{f}$

the closure is disconnected

$g([\mathrm{D}\mathrm{G}])$

In the case of the exponential family, Devaney and Goldber tained the explicit expression
$\varphi^{-1}\circ E_{\lambda\varphi(z}\circ)=\exp i(\frac{\mu+\overline{\mu}z}{1+z})$

ob-

$\mu\in\{z|1\mathrm{m}z>0\}$

for a suitable Riemann map which was crucial to show the density of in . In general, of course, we cannot obtain the explicit form of so instead of it we take advantage of a property of inner is called an inner functions. In general analytic function : exists for almost every function if the radial limit and satises $|g(e^{i\theta})|=1$ . It is easy to see that is an inner function. It is known that an inner function has a unique tends to locally xed point.. called a Denjoy-Wol point and . The following is an important lemma for the uniformly on proof of the Main Theorem.
$\varphi$

$\Theta_{\infty}$

$\partial \mathrm{D}$

$\varphi^{-1}\mathrm{o}f^{n0}\mathrm{o}\varphi(z)$

$g$

$\mathrm{D}arrow \mathrm{D}$

$g(e^{i\theta}):= \lim_{r\nearrow 1g}(re^{i})\theta$

$e^{i\theta}\in\partial \mathrm{D}$

$\varphi^{-1}\circ f^{n_{0}}\circ\varphi$

$g$

$p\in\overline{\mathrm{D}}$

$g^{n}(z)$

$p$

$\mathrm{D}([\mathrm{D}\mathrm{M}])$

be an inner function which is not a M\"obius transformation and its Denjoy-Wol point. where $E$ holds for every (1) If then is a certain exceptional set of logarithmic capacity zero. where $E$ holds for every then (2) If capacity zero and $K$ is a certain is a certain exceptional set of loga . perfect set in

Lemma 1 Let
$p\in \mathrm{D},$

$g$

$\mathrm{D}arrow \mathrm{D}$

$p$

$\overline{\bigcup_{n}^{\infty}=1g^{-n}(z_{0})}\supset\partial \mathrm{D}$

$z_{0}\in \mathrm{D}\backslash E$

$p\in\partial \mathrm{D}_{f}$

$\overline{\bigcup_{n1}^{\infty-n}=g(Z\mathrm{o})}\supset K$

$z_{0}\in \mathrm{D}\backslash E$

$7\dot{\mathrm{v}}thmic$

$\partial \mathrm{D}$

117

is either an attracting basin or a parabolic basin and . we can say more about the set If
$U$
$\bigcup_{n=1}^{\infty-n}g(Z0)$

$g=\varphi^{-1}\circ f^{n_{0}}\circ\varphi$

Lemma 2 Let $U$ be either an attracting basin or a parabolic basin (not . Then there exists a set necessarily unbounded) and capacity zero such that of loga
$g=\varphi^{-1}\mathrm{o}f^{n_{0}}\circ\varphi$
$E\subset \mathrm{D}$ $r\dot{\mathrm{v}}thmic$
$\frac{\sigma_{n}(z_{0},A)}{\sigma_{n}(z_{0},\partial \mathrm{D})}arrow\frac{measA}{2\pi}$

$(narrow\infty)$

holds

for every
.

$\sum(1-|\zeta|^{2})$

and every arc and sum is taken over all

$z_{0}\in \mathrm{D}\backslash E$

$A$

in

$\partial \mathrm{D}_{f}$

$\zeta=|\zeta|e^{i\theta}$

where $\sigma_{n}(z_{0}, A)=$ with $g^{n}(\zeta)=z_{0}$ and

$e^{i\theta}\in A\zeta$

The conclusion of Lemma 2 is stronger than that of Lemma 1 (1), because accumulate on all over it implies not only that the inverse images . We shall not give but also that their distribution is uniform on the denition of logarithmic capacity here (see [P2]). But we recall that a set of logarithmic capacity zero is extremely thin: it cannot contain a connected set with more than one point and its Haus,d or dimension is zero ([DM], [P2]). \S 2 we prove Theorem 1 and Corolary 1. \S 3 consists of three subsections. In \S 3.1 we prove Theorem 2 and make some remarks on the sucient conditions for $f$ to admit no unbounded Fatou components. \S 3.2 we prove Lemma 1 and Lemma 2 which are keys for the proof of the Main Theorem. In \S 3.3 we prove the Main Theorem.
$g^{-n}(z\mathrm{o})$
$\partial \mathrm{D}$ $\partial \mathrm{D}$

$\ln$

$\ln$

Connectivity of

$J_{f}\cup\{\infty\}$

in

$\overline{\mathbb{C}}$

(Proof of Theorem 1): The following criterion is well known. (See for example [Bea], p.81, Proposition 5.1.5).

Proposition Let $K$ be a compact subset in . Then $K$ is connected is simply connected. and only if each component of the complement
$\overline{\mathbb{C}}$

$\mathrm{B}$

if

$K^{c}$

is compact in , we can apply Proposition B. As we Since mentioned in \S 1, eventually periodic components are simply connected. So if a Fatou component is not simply connected, then is necessarily
$\overline{\mathbb{C}}$

$J_{f}\cup\{\infty\}$

$U$

$U$

118

$\mathrm{p}\mathrm{r}\mathrm{o}\mathrm{O}\mathrm{a}\mathrm{w}\mathrm{a}\mathrm{n}\mathrm{f}$

.dering domain which is not simply connected.

$F_{f}([\mathrm{E}\mathrm{L}2])$ $f^{n}$

This completes

$\mathrm{t}\mathrm{h}\mathrm{e}$

(Proof of Corollary 1): Under the condition (1), cannot tend to through . On the other hand, tends to on any multiply-connected wandering domains . So all the Fatou components are simply connected in this case. Under the condition (2) or (3), it is known that all the Fatou components must be simply ([Ba4], [EL1], p.620 Corollary 1, 2).
$f^{n}$
$\infty$

$\infty$

$([\mathrm{B}\mathrm{a}4], [\mathrm{E}\mathrm{L}1])$

$\mathrm{c}\mathrm{o}\mathrm{n}\mathrm{n}\mathrm{e}\mathrm{C}\mathrm{t}\mathrm{e}\mathrm{d}\square$

Remark 1 (1) Let . Then there is even no wandering domain in for . For $f\in B,$ may admit a wandering domain but must be simply connected as we mentioned above. Under an additional condition
$S:=\{f|\#_{\mathrm{s}\mathrm{i}\mathrm{n}\mathrm{g}(f)}-1<\infty\}\subset B$
$F_{f}$

$f\in S([\mathrm{G}\mathrm{K}])$

$F_{f}$

$U$

$U$

$J_{f}\cap(\mathrm{d}\mathrm{e}\mathrm{r}\mathrm{i}_{\mathrm{V}}\mathrm{e}\mathrm{d}$

set of

$\bigcup_{n=0}^{\infty}fn(\mathrm{s}\mathrm{i}\mathrm{n}\mathrm{g}(f^{-}1)))=\emptyset$

has also no wandering domain ([BHKMT]). (2) We can conclude that in general if is disconnected, all the Fatou components are bounded and some of which are multiply-connected wandering domains.
$f\in B$
$J_{f}\cup\{\infty\}$

3
3.1

Connectivity of

$J_{f}$

in

$\mathbb{C}$

The case when all the Fatou components are bounded

$\mathbb{C}$

Suppose that a closed connected subset $K$ in is bounded. Then all the components of the complement other than the unique unbounded $V$ component are simply connceted. (Of course, is simply $K$ connected). If is unbounded, then all the components of are simply connected, but the converse is false as the example shows. (Compare with the Proposition B). But note that is connected in . For the connectivity of a closed subset in , the following criterion holds.
$K^{c}$
$V\cup\{\infty\}\subset\overline{\mathbb{C}}$

$K^{c}$

$J_{F_{\lrcorner\lambda}}(0< \lambda<\frac{1}{e})$

$J_{E_{\lambda}}\cup\{\infty\}$

$\overline{\mathbb{C}}$

$\mathbb{C}$

Proposition 1 Let $K$ be a closed subset of C. Then $K$ is connected if and only if the boundary of each component is of the complement connected.
$U$
$K^{c}$

119

(Proof): For the only if part, see [New]. Suppose that $K$ is disconnected. Then there exist two closed sets and with $K=K_{1}\cup K_{2}$ and . Take a point with $d(z_{0,1}K)=d(z0, K2)$ where denotes the Euclid distance in C. Then $z_{0}\in K^{c}$ and so let be the connected component of containing . Since is connected by the assumption, either or . Without loss of generality we can assume . On the other hand denote $r_{0}:=d(z_{0,1}K)=d(z_{0}, K_{2})$ and let $D_{r_{0}}(\mathcal{Z}_{0}):=\{z||z-z\mathrm{o}|<r_{0}\}.$ Then and there exists a point $w\in K_{2}$ with $w \in K_{2}\subset K$ . Since , we have but this is a contradiction since and . This completes the .
$K_{1}$
$K_{2}$

$K_{1}\cap K_{2}=\emptyset$

$z_{0}$

$d$

$U_{0}$

$K^{c}$

$z_{0}$

$\partial U_{0}$

$\partial U_{0}\subset K_{1}$

$\partial U_{0}\subset K_{2}$

$\partial U_{0}\subset K_{1}$

$\overline{D_{r_{0}}(Z_{0})}\subset\overline{U_{0}}$

$w\in\overline{U_{0}}$

$w\in\partial U_{0}$

$\partial U_{0}\subset K_{1}$

$K_{1}\cap K_{2}=\emptyset$

$\mathrm{p}\mathrm{r}\mathrm{o}\mathrm{o}\mathrm{f}\square$

(Proof of Theorem 2): By Proposition 1, it is sucient to to show that the boundary is connected for each Fatou component . Since is bounded, the boundary of as a subset of and the one as the subset coincide. Hence $U$ is simply connected if and only if is ([Bea], p.81, Proposition 5.1.4). This completes the proof.
$\partial U$

$U$

$U$

$U$

$\mathbb{C}$

$\mathrm{o}\mathrm{f}\overline{\mathbb{C}}$

$\partial U$

$\mathrm{c}\mathrm{o}\mathrm{n}\mathrm{n}\mathrm{e}\mathrm{C}\mathrm{t}\mathrm{e}\mathrm{d}\square$

Remark 2 (1) Since a non-simply connected Fatou component is necessarily a wandering domain, the assumption of Theorem 2 is equivalent to that the components of are bounded and admits no multiplyconnected wandering domains. (2) Several sucient conditions are known for a transcendental entire function to admit no unbounded Fatou components as follows: (i) $([\mathrm{B}\mathrm{a}3])\log M(r)=O((\log r.)^{p})$ (as $rarrow\infty$ ) where
$\mathrm{a}\mathbb{I}$

$F_{f}$

$F_{f}$

$f$

$M(r)=|z|=r\mathrm{s}\mathrm{u}\mathrm{p}|f(Z)|$

and $1<p<3$ .
(ii)
$([\mathrm{S}])$

There exists

$\epsilon\in(0,1)$

such that log log

$M(r)< \frac{(\log r)^{\frac{1}{2}}}{(\log\log r)^{\epsilon}}$

for

large
(iii)

$r$

.
The order of is less than
$f$
$\frac{1}{2}$

$([\mathrm{S}])$

and

$\frac{\log M(2r)}{\log M(r)}arrow c$

(nite constant)

as . Note that the condition (ii) includes the condition (i).

$rarrow\infty$

3.2

A property of inner functions

$g$
$g$

(Proof of Lemma 1): If is a nite Blaschke product, then is a rational function of degree $d\geq 2$ . It is well known that in general the

120

$p$

closure of the set of all the inverse images of by a rational function for any which is not a Fatou contains its Julia set exceptional point ([Bea], p.79, Theorem 4.2.7). If the Denjoy-Wol point , then is in , then and if the Denjoy-Wol point is in or at least . In any cases all the inverse is a perfect set in . So our assertion holds for images of are in for every
$z_{0}$

$R$

$\mathrm{o}\mathrm{f}_{\text{}}\mathrm{d}\mathrm{e}\mathrm{g}\mathrm{r}\mathrm{e}\mathrm{e}d\geq 2$

$J_{R}$

$z_{0}$

$\mathrm{D}$

$J_{g}=\partial \mathrm{D}$

$p$

$\partial \mathrm{D}$

$J_{g}=\partial \mathrm{D}$

$J_{g}$

$\partial \mathrm{D}$

$\mathrm{D}$

$z_{0}$

$z_{0}\in \mathrm{D}$

$E=E(g):=$ and we have

$g$

{ $z|z$ is a Fatou exceptional point for }

$g$

$\#_{E(g)}\leq 2$

, which implies that

$E$

is a set of

$1\mathrm{o}g$

arithmic

$T_{a}\circ g$

capacity zero. lf is not a nite Blaschke product, then by Frostmans theorem ([G], p.79, Theorem 6.4) there exists a set of capacity zero such that $z-a$ is a Blaschke product for every , where $T_{a}(z):=-$ .
$E_{1}\in \mathrm{D}$
$a\in \mathrm{D}\backslash E_{1}$

$1-\overline{a}z$

Therefore $B:=T_{a^{\circ}}g\circ Ta-1$ is also a Blaschke product. By applying Frostinans theorem to each , we obtain the set of logarithmic is a Blaschke holds and each capacity zero such that product for every . Now it is sucient to prove our lemma and corresponding for , so we concentrate on a xed go . Let Blaschke product be the set of accumulation $B ^{-n}(0)$ , then points of is closed and can be analytically continued to a meromorphic function on by the reection principle ([G], p.75, is equal to the set of singularities of Theorem 6.1). In other words, (that is, points at which $B^{n}(z)$ does not extend analytically). There exists a set of logarithmic capacity zero such that is equal to the set of , Theorem 6.6). Let ( accumulation points of $B^{-n}(p)$ for , then $E$ is a set of capacity zero and for every . we have First let us consider the case when the Denjoy-Wol point is in D. , then there Suppose that does not hold for a can be dened on such that with exists a open set . We take as the maximal for every and . Since $B$ is not a nite set satisfying this property. Let . Hence for a Blaschke product, we have $\#\{B^{-1}(0)\}=\infty$ and so . Since . So there exists a point we have is a normal family. Then and cannot take the values by the dynamics of $B$ on , we have $B^{n}|Varrow p$ locally uniformly. But on by the construction of the extension, the other hand
$g^{n}$
$\bigcup_{n1}^{\infty}=E_{i}$

$\tau_{a}\mathrm{o}g^{n_{\mathrm{O}}}\tau a-1=B^{n}$

$B^{n}$

$a \in \mathrm{D}\backslash (\bigcup_{n=1}^{\infty}Ei)$

$B$

$a \in \mathrm{D}\backslash (\bigcup_{n=1i}^{\infty}E)$

$B=T_{a}\mathrm{o}$
$A_{n}$

$\tau_{a}-1$

$A_{n}\subset\partial \mathrm{D}$

$B^{n}$

$\overline{\mathbb{C}}\backslash A_{n}$

$A_{n}$

$B^{n}$

$E_{B_{n}}$

$A_{n}$

$p\in \mathrm{D}\backslash E_{B_{n}}$

$[\mathrm{G}]$

$E:= \bigcup_{n=1}^{\infty}EB_{n}$

$z_{0}\in \mathrm{D}\backslash E$

$\bigcup_{n=1}^{\infty}B^{-n}(z\mathrm{o})\supset\overline{\bigcup_{n1}^{\infty}=A_{n}}$

$p$

$\overline{\bigcup_{n=1}^{\infty}B^{-n}(Z\mathrm{o})}\supset\partial \mathrm{D}$

$z_{0}\in \mathrm{D}\backslash E$

$V$

$V\cap\partial \mathrm{D}\neq\emptyset$

$B^{n}$

$V$

$n\in \mathrm{N}$

$V \cap(\bigcup_{n=1}^{\infty}B^{-}n(z_{0}))=\emptyset$

$V$

$W:=V\cap\partial \mathrm{D}$

$A_{1}\neq\emptyset$

$z_{0}\in \mathrm{D}\backslash E$

$W\neq\partial \mathrm{D}$

$\alpha\in\partial \mathrm{D}\backslash W$

$B^{n}$

$z_{0},\overline{\overline{z_{0}}}1$

$\alpha,$

$\{B^{n}|V\}^{\infty}n=1$

$\mathrm{D}$

$B^{n}|(V\cap(\overline{\mathrm{D}})^{C})arrow\overline{\overline{p}}1$

121

holds in this case. which is a contradiction. Hence Next we consider the case when the Denjoy-Wol point is on the . Then and suppose that boundary of D. Let is dened on for every $n\in$ N. Obviously $K$ is closed. If $K$ and consists of a single point, say , then we have $B$ is one to one since is extended by the reection principle. It follows that $B$ is a M\"obius transformation, which is a contradiction. By the similar argument, we can prove $\#_{K}\geq 3$ . Then $K$ cannot have an isolated point. If this is not the case, let $\beta\in K$ be an isolated point. hence by Picards theorem, $B$ takes Then is an essential singularity two values in innitely often. This contradicts the all but and $\#_{K}\geq 3$ . Therefore it follows that $K$ fact that , this holds for every is a perfect set: Since completes the proof. (Proof of Lemma 2): In the case when $U$ is an attracting basin, the is a result is a special case of Theorem 3 in [P1]. In the case when parabolic basin, the result -follows by combining the series of theorems in [DM] (Theorem 6.1, Theorem 4.2, Corollary 4.3, Theorem 3.1) with the Theorem 3 in [P1]. .
$\bigcup_{n=1}^{\infty}B^{-n}(z\mathrm{o})\supset\partial \mathrm{D}$

$p$

$K:= \bigcup_{n=1}^{\infty}A_{n}$

$K\neq\partial \mathrm{D}$

$B^{n}$

$\overline{C}\backslash K$

$\beta$

$B(\partial \mathrm{D}\backslash \{\beta\})\subset\partial \mathrm{D}\backslash \{\beta\}$

$B|(\partial \mathrm{D}\backslash \{\beta\})$

$\beta$

$\underline{\mathrm{a}}\mathrm{n}\mathrm{d}$

$\mathbb{C}$

$\mathrm{e}\mathrm{X}\mathrm{C}\mathrm{e}_{\underline{\mathrm{P}^{\mathrm{t}\mathrm{i}\mathrm{n}\mathrm{a}}}}\mathrm{o}1$

$B(\mathbb{C}\backslash K)\subset\overline{\mathbb{C}}\backslash K$

$\overline{\cup.n=1B^{-n}\infty(z\mathrm{o})}\supset K$

$z_{0}\in \mathrm{D}\backslash E$

$\square$

$U$

$\mathrm{t}\mathrm{o}\mathrm{g}\mathrm{e}\mathrm{t}\mathrm{h}\mathrm{e}\coprod \mathrm{r}$

$t$

3.3

admits an unbounded component On In the case when the Boundary of unbounded invariant Fatou Components
$F_{f}$

(Proof of Main Theorem): In what follows we assume that $n_{0}=1$ (that is, is an invariant component) and $m_{0}=1$ for simplicity. This instead causes no loss of generality, because we have only to consider of in general cases. Case (1) Since is accessible, there exists a continuous curve $L(t)(0\leq$ with $\lim_{tarrow 1}L(t)=\infty$ . By deforming $L(t)$ slightly, we construct $t<1)$ in a new curve satisfying the following condition.
$U$
$f^{m_{0}}$

$f$

$\infty$

$U$

$\mathcal{L}(t)$

Lemma 3 There exists a curve $\mathcal{L}(t)(0\leq t<1)$ with can be analytically continued along it such that every branch of . every
$f^{-n}$
$n\in \mathrm{N}$

$\lim_{tarrow 1}\mathcal{L}(t)=\infty$

for

. we have (Proof): We may assume that $L(0)\not\in P_{f}$ , since be points on such that all the piecewise linear Let $p_{0:=}L(\mathrm{o}),p_{1},p2,$ , lie in . Let line segments connecting
$q\not\in P_{f}$
$U\not\subset P_{f}$
$\ldots$

$L$

$p_{0},p_{1,p_{2}},$

$\ldots$

$U$

$F_{n}^{(1)},$

$F_{n}(2),$

$\ldots$

$F_{n}^{(m)},$

$\ldots$

122

which take values on be all the branches of $m$ may be nite or innite. Dene
$f^{-n}$
$\Theta_{n}^{(m)}(p_{0}):=\{e^{i\theta}|F_{n}^{(m)}\mathrm{c}\mathrm{a}\mathrm{n}$

$U$

. The range of the sux

$|\mathrm{h}\mathrm{o}\mathrm{m}p_{0}$

be analyticaly continued along the ray in the direction } $(n=1,2, \ldots)$ .

$\theta$

Then by the next Grosss Star Theorem . has full measure in

$\partial \mathrm{D}$
$\mathrm{C}$

$([\mathrm{N}\mathrm{e}\mathrm{v}])$

, it follows that

$\Theta_{n}^{(m)}(p_{0})$

(Grosss Star Theorem) Let be an entire function and Lemma . Then $F$ can be a branch of dened in the neighborhood of in the direction . analytically continued along almost all rays from
$f$
$F$
$f^{-1}$
$p_{0}\in \mathbb{C}$
$\theta$

$p_{0}$

Then the set

$\Theta(p\mathrm{o}):=\bigcap_{n\geq 1,m\geq 1}\Theta_{n}(m)(p_{0})$

. Hence by changing slightly to a point , the has also full measure in and lie in $U$ and all the branches $F_{n}^{(m)}(n\geq 1, m\geq 1)$ segments . By the same method, we can can be analytically continued along lies in $U$ and has the nd a point close to such that the segment same property as above. By repeating this argument, we can prove Lemma 3.
$\partial \mathrm{D}$

$p_{1}$

$p_{1}$

$\overline{p_{0}p_{1}}$

$\overline{p_{1}p_{2}}$

$\overline{p_{0}p_{1}}$

$p_{2}$

$\overline{p_{1p_{2}}^{\prime/}}$

$p_{2}$

$\mathrm{t}\mathrm{h}\mathrm{e}\square$

. For supthen we have pose this is false, then there exist an increasing sequence of parameter values $t_{1}<t_{2}<\cdots<t_{k}<\cdots$ and a nite point with and this contra. Then it follows that . dicts the fact be a Riemann map of . Then Let :
Let
$l_{n}^{(m)}(t):=F_{n}^{(m)}(\mathcal{L}(t))$

$\lim_{tarrow 1}l_{n}^{(m)}(t)=\infty$

$\alpha$

$\lim_{karrow\infty}\iota_{n}^{(m)}(\mathrm{t}_{k})=$

$\alpha\neq\infty$

$\lim_{karrow\infty}\mathcal{L}(t_{k})=f^{n}(\alpha)\neq\infty$

$\lim_{karrow\infty}\mathcal{L}(t_{k})=\infty$

$\varphi$

$\mathrm{D}\mapsto U$

$U$

$\Gamma(t):=\varphi^{-1}(\mathcal{L}(t))$

and

$\gamma_{n}^{(m)}(t):=\varphi^{-1}(l_{n}^{(m})(t))$

. This fact is not so trivial but landing at a point in follows from the proposition in [P2](p.29, Proposition 2.14). We may for simplicity. If lands at assume that , it follows that , then since and this is equal to . This there exists the radial limit the theorem in [P2] (p.34, Theorem 2.16). Therefore it fact follows is sucient to show that the set of all the landing points of $m\geq 1)$ is dense in .
are curves in
$\mathrm{D}$
$\partial \mathrm{D}$

$\Gamma(t)$

$z=1\in\partial \mathrm{D}$

$\lim_{tarrow 1\gamma n}(m0)(0t)=$

$e^{i\theta_{0}}$

$\lim_{tarrow 1\varphi}(\gamma_{n_{0}}^{(}(m_{0})t))=\lim_{tarrow 1}l_{n_{0}}(m_{0})(t)=\infty$

$\lim_{rarrow 1}\varphi(rei\theta_{0})$

$\infty$

$\mathrm{h}\mathrm{o}\mathrm{m}$

$\gamma_{n}^{(m)}(t)(n\geq$

$1,$

$\partial \mathrm{D}$

123

. Then by Fatous theorem has radial : Let limit and non-constant for almost every for almost . Hence is a curve landing at a point in . every . Therefore it follows that and thus is an inner function. Let then by the same reason . for is a curve in with an end point ing
$g:=\varphi^{-1}\mathrm{o}f\mathrm{o}\varphi$
$\mathrm{D}arrow \mathrm{D}$
$\varphi$

$\varphi(e^{i\theta})=\lim_{r\nearrow 1\varphi}(re^{i})\theta\in\partial U$

$e^{i\theta}\in$

$\partial \mathrm{D}$

$f\mathrm{o}\varphi(re^{i})\theta$

$\partial U\backslash \{\infty\}$

$e^{i\theta}\in\partial \mathrm{D}$

$\lim_{r\nearrow 1\varphi^{-1}}\circ f\circ\varphi(re^{i})\theta\in\partial \mathrm{D}\mathrm{a}.\mathrm{e}$

$g$

$\overline{C}:=\varphi^{-1}(C)$

$\Gamma(t),\overline{C}$

$\mathrm{D}$

$g(\overline{C})\supset\overline{C}$

$\overline{q}\in\partial U\mathrm{s}\mathrm{a}\mathrm{t}\mathrm{i}\mathrm{S}\mathfrak{h}\infty r$

Rom the dynamics of

$\overline{\mathrm{D}}$

$g:\mathrm{D}arrow \mathrm{D}$

, it follows that the set

$\overline{p}$

$\bigcup_{n=0}g^{n}(\overline{c})\cup\{\overline{p},\overline{q}\}$

$\epsilon_{0}>0$

and is an attracting xed point of is compact in where and the distance between this set and $z=1$ is positive. Hence there exists such that
$\overline{p}=\varphi^{-1}(p)$

$g$

$U_{\epsilon_{0}}(1) \cap\{\bigcup_{0=}gn(n\overline{C})\cup\infty\{\overline{p},\overline{q}\}\}=\emptyset$

(1)
$\Gamma|[t_{0},1)\subset$

lands at $z=1$ , there exists $t_{0}\in[0,1)$ such that . So by rewriting to $\Gamma(t)(0\leq t<1)$ we may assume that for $0\leq t<1$ ). Let $K:=\{z||z|\leq 1-\epsilon_{0}\}$ then since every such and $K$ is compact, there exists point in tends to under . Then we have that for every $N\geq n_{1}$ we have , then by operating for every $N\geq n_{1}$ . For suppose that . which contradicts we have Now suppose that the conclusion does not hold. Then there exists

Since

$\Gamma(t)$

$U_{\epsilon_{0}}(1)$

$\Gamma|[t_{0},1)$

$\Gamma(t)\subset U_{\epsilon_{0}}(1)$

$\mathrm{D}$

$\overline{p}$

$g^{n}$

$n_{1}\in \mathrm{N}$

$g^{N}(K)\subset U_{\epsilon}(\overline{p})$

$\gamma_{N}^{(m)}(t)\subset K^{c}$

$\gamma_{N}^{(m)}(t)\cap K\neq\emptyset$

$f^{N}$

$\Gamma(t)\cap K\neq\emptyset$

$\Gamma(t)\subset U_{\epsilon_{0}}(1)$

$(\theta_{1}, \theta_{2}):=\{e^{i\theta}|\theta_{1}<\theta<\theta_{2}\}\subset\partial \mathrm{D}$

with

$\Theta_{\infty}\cap(\theta_{1}, \theta_{2})=\emptyset$

.
$\partial \mathrm{D}$

$\gamma_{n_{1}}^{(m_{1})}(t)\subset K^{c}$

By changing the starting point slightly, if necessary, we may assume that the points $\gamma_{n}^{(m)}(0)(n, m=1,2, \cdots)$ accumulate to all over $m=$ by Lemma 1 (1) while the end points $1,2,$ such that . Therefore there exists are not in and which have limit On the other hand there exist inverse images , there densely. The reason is as follows: Since points on
$\Gamma(0)$

$\gamma_{n}^{(m)}(1):=\lim_{tarrow 1}\gamma_{n}^{(}(m)t)(n,$
$\gamma_{n_{1}}^{(m_{1})}(t)$

$\cdots)$

$(\theta_{1}, \theta_{2})$

$\gamma_{n_{1}}^{(m_{1})}(1)\in\partial \mathrm{D}\backslash (\theta_{1}, \theta_{2})$

$g^{-n}(\overline{C})$

$(\theta_{1}, \theta_{2})$

$q\not\in P_{f}$

$\mathrm{e}..\mathrm{x}.\mathrm{i}\mathrm{S}\mathrm{t}_{\mathrm{S}}\mathrm{a}\mathrm{n}\mathrm{c}\mathrm{a}\mathrm{n}\mathrm{b}\mathrm{e}\mathrm{d}\mathrm{e}\mathrm{i}\mathrm{g}\mathrm{h}\mathrm{b}_{\mathrm{o}\mathrm{r}}\mathrm{h}\mathrm{o}\mathrm{o}_{\mathrm{L}}\mathrm{d}V\mathrm{o}\mathrm{f}\mathrm{e}\mathrm{f}\mathrm{i}\mathrm{n}\mathrm{e}\mathrm{d}.\mathrm{e}\mathrm{t}V0\subset V\mathrm{i}\mathrm{S}\mathrm{a}\mathrm{n}\mathrm{e}\mathrm{i}q\mathrm{S}\mathrm{u}\mathrm{c}\mathrm{h}\mathrm{t}\mathrm{h}\mathrm{a}\mathrm{t}\mathrm{a}\mathrm{n}_{\mathrm{h}}\mathrm{t}\mathrm{h}\mathrm{e}\mathrm{b}\mathrm{r}_{\mathrm{O}\mathrm{o}\mathrm{d}_{0}}\mathrm{a}\mathrm{n}\mathrm{c}\mathrm{h}\mathrm{e}\mathrm{s}F^{(1)},F\ldots,$

$F^{(m}) \mathrm{g}\mathrm{b}\mathrm{o}\mathrm{r}\mathrm{h}\mathrm{f}q\mathrm{W}\mathrm{h}\frac{n(2)}{V_{0}}\subset V.\mathrm{w}\mathrm{e}n_{\mathrm{i}\mathrm{t}}n$

may assume that

Then

$C\subset V_{0}$

. Dene
,
$\partial \mathrm{D}$

$c_{n}^{(m)}(t):=F_{n}^{()}m(C(t))$

$\overline{c}_{n}^{(m)}(t):=\varphi^{-}(1c^{()}(nmt))$

.
$\overline{c}_{n}^{(m)}(t)$

$c_{n}^{(m)}(t)$

curve in

$\mathrm{D}$

is a and landing at a point in is a curve in by the same reason as before. Let landing at a point in
$U$
$\partial U$

124

. By changing the starting point be any subarc of $m=$ slightly, if necessary, we may assume that the points $1,2,$ by Lemma 1 (1). Since radial limits of accumulate to and slightly exist and non-constant almost everywhere, by changing if necessary, we may assume that there exist the nite values accumulate on . Then and with . In general the family of single-valued is normal and furtheranalytic branch of $f^{-n}(n=1,2, \ldots)$ on a domain , any local uniform limit of a subsequence in the family more if is constant ([Bea], p.193, Theorem 9.2.1, Lemma 9.2.2). So the family is normal and all its limit functions are constant and hence for a tends to zero, that is, suitable subsequence the diameter of if the conmust land at a point in . If the must land at a point in stant limit is nite. Therefore cannot intersect the curves constant limit is , for large enough both $\{\varphi(re^{i\theta}3)|0\leq r\leq 1\}$ and $\{\varphi(re^{i\theta}4)|0\leq r\leq 1\}$ which are bounded set, since the convergence is uniform on . Hence ain we can conclude must land at a point in that . This proves the asmust land at a point in and therefore sertion. . We may assuch that Then there exists , we have then since sume that $n_{1}>N_{1}$ . Let and $n_{1}>N_{1}$ , we have . On the other hand since . is dense in which contradicts (1). Therefore easily follows by the same argument as in the Disconnectivity of of in \S 1. This completes the proof in the case of (1).
$(\underline{\theta}_{3}, \theta_{4})\subset(\theta_{1}, \theta_{2})$

$(\theta_{1}, \theta_{2})$

$C(\mathrm{O})$

$\overline{c}_{n}^{(m)}(0)(n,$

$\cdots)$

$(\theta_{3}, \theta_{4})$

$\varphi$

$\theta_{3}$

$\theta_{4}$

$\varphi(e^{i\theta_{3}})$

$\varphi(e^{i\theta_{4}})$

$\varphi(e^{i\theta_{3}})\neq\varphi(e^{i\theta_{4}})$

$c_{n}^{(m)}(0)$

$\partial U\cap$

$\varphi(\{re^{i\theta}|\theta_{3}<\theta<\theta_{4},0\leq r\leq 1\})$

$U_{0}$

$U_{0}\cap J_{f}\neq\emptyset$

$\{F_{n}^{(m)}|V\mathrm{o}\}$

$c_{n_{k}}(m_{k})(t)$

$c_{n_{k}}(m_{k})(t)$

$\partial U\cap\varphi(\{re^{i\theta}|\theta_{3}<\theta<\theta_{4},0\leq r\leq 1\})$

$\overline{C}_{n_{k}}^{(m_{k})}(t)$

$(\theta_{3}, \theta_{4})$

$\infty$

$n_{k}$

$c_{nk}^{(m_{k})}$

$V_{0}$

$\mathrm{a}g$

$C_{n_{k}}^{(m_{k})}(t)$

$\partial U\cap\varphi(\{re^{i}\theta|\theta_{3}<\theta<\theta_{4},0\leq r\leq 1\})$

$\overline{c_{n_{k}}.}(m_{k})(t)$

$(\theta_{3}, \theta_{4})$

$\overline{c}_{N1}^{(M_{1})}$

$\gamma_{n_{1}}^{(m_{1})}\cap\overline{c}_{N1}^{(M_{1})}\neq\emptyset$

$u\in\gamma_{n_{1}}^{(m)}1\cap\overline{c}_{N1}^{(M_{1})}$

$u\in\gamma_{n_{1}}^{(m)}1$

$g^{n_{1}}(u)\in U_{\epsilon_{0}}(1)$

$u\in\overline{c}_{N1}^{(M_{1})}$

$g^{n_{1}}(u) \in\bigcup_{n=0}^{\infty}g^{n}(\overline{c})$

$\Theta_{\infty}$

$\partial \mathrm{D}$

$J_{f}$

$\mathrm{c}\mathrm{a}\mathrm{s}\mathrm{e}\square$

$E_{\lambda}$

$\mathrm{C}\underline{\mathrm{a}\mathrm{s}\mathrm{e}(2)}$

2,

$\bigcup_{n=1}^{\infty}g^{-n}(Z0)\supset\partial \mathrm{D}(z_{0}\in \mathrm{D}\backslash E)$

The proof is quite parallel to the case (1). Note that by Lemma in this case. holds for
$g=\varphi^{-1}\mathrm{o}f\mathrm{o}\varphi$
$\square$

with Since is unique and denote it by by and are dense in points of $g(z)$ is an irrational rotation and
$g(z)=e^{2\pi i\theta_{0}}$
$\mathrm{C}\mathrm{a}\mathrm{s}\mathrm{e}(3)-n$

$\theta_{0}\in \mathbb{R}\backslash \mathbb{Q}$

$g$

$\gamma_{n}(t)$

$\gamma_{n}(t)$

$\partial \mathrm{D}$

$\varphi$

, the inverse image of . Then it is obvious that the end there, since attains radial limit
$\Gamma(t)$
$\infty$

$\lim_{tarrow 1}\varphi(\gamma n(t))=\lim f-1(tarrow 1\varphi(\Gamma(t)))=\infty$

.
$\infty$

$\square$

this case we need not assume the accessibility of Case (4) . The set this condition is automaticaly satised
$\ln$
$([\mathrm{B}\mathrm{a}6])$

$\bigcup_{n=0}^{\infty}fn(C)$

, because is a

125

-intersections and tends to . It is not dicult . Hence we have . The rest of the proof is quite parallel to the case (1) if the conclusion of Lemma 2 (1) holds for . If we have only the conclusion of Lemma 2 (2), then we can prove that for every arc with holds by the similar argument.
$\mathrm{s}\mathrm{e}\mathrm{l}\dot{\mathrm{f}}$

curve which may have to take satisfying

$L$
$\emptyset$

$\infty$

$L \cap(\bigcup_{n0}^{\infty}f=n(c))=\emptyset$

$\mathcal{L}\cap(\bigcup_{n0}^{\infty}f=n(c))=$

$g$

$A\subset\partial \mathrm{D}$

$A\cap K\neq\emptyset,$

$A\cap\Theta_{\infty}\neq\emptyset\square$

References
[Bal] I. N. Baker, Multiply connected domains theory, Math. Z. 81, (1963), 206-214.

of normality

in iteration

[Ba2] I. N. Baker, An entire function which has a wandering domain, J. Austral. Math. Soc Ser. A 22, (1976), 173-176.
[Ba3] I. N. Baker, The iteration of polynomials and transcendental entire functions, J. Austral. Math. Soc. Ser. A 30 (1981), 483-495.

[Ba4] I. N. Baker, Wandering domains in the iteration Proc. London Math. Soc.(3) 49, (1984), 563-576.

of entire functions,

[Ba5] I. N. Baker, Some entire functions with multiply-connected wandering domains, Erg. Th. &Dyn. Sys. 5 (1985), 163-169.

[Ba6] I. N. Baker, Innite limits in the iterations god. Th. &Dynam. Sys. 8 (1988), 503-507.

of entire functions,

Er-

[Bea] A. F. Beardon, Iteration of Rational Functions, Springer Verlag, New York, Berlin and Heiderberg, 1991.
[Ber] W. Bergweiler, Iteration (1993) No.2, 151-188.

of meromorphic functions,

Bull. AMS, 29

[BHKMT] W. Bergweiler, M. Haruta, H. Kriete, H.-G. Meier and N. Terglance, On the limit functions of iterates in wandering domains, Ann. Acad. Sci. Fenn. Ser. A I Math 18 (1993), 369-375.
[D] R. L. Devaney, : Dynamics and bifurcations, Internat. J. Bif. & Chaos, 1 (1991), 287-308.
$e^{z}$

[DG] R. L. Devaney and L. R. Goldberg,

Uniformization of

attracting

126

basins 266.

for

exponential maps, Duke. Math. J., 55 No.2 (1987), 253$\mathrm{M}\mathrm{a}\tilde{\mathrm{n}}\acute{\mathrm{e}}$

[DM] C. I. Doering and R. , The Dynamics Ensaios Matem\aticos 3 (1991).

of Inner

Functions,

[EL1] A. E. Eremenko and M. Yu. Lyubich, The dynamics transformations, Leningrad Math. J. 1 (1990), 563-634

of

analytic

[EL2] A. E. Eremenko and M. Yu. Lyubich, Dynamical properties of some classes of entire functions, Ann. lnst. Fourier (Grenoble) 42 (1992),

989-1020.

[G] J. B. Garnett, Bounded analytic functions, Academic Press, New York, (1981). [GK] L. R. Goldberg and L. Keen, A niteness theorem for a dynamical class of entire functions, Ergod. Th. &Dynam. Sys. 6 (1986), 183-192. [Ki] M. Kisaka, Local uniform convergence and convergence Nonlinearity 8 (1995) 273-281.
$\lambda(1+\frac{z}{d})^{d}$

of Julia sets,
of
$\lambda e^{z}$

[Kr] B. Krauskopf, Convergence of Julia sets in the approximation by , Internat. J. Bif. &Chaos, 3 (1993), 257-270.

[KrK] B. Krauskopf and H. Kriete, Preprint.

Hausdor convergence of Julia

sets,

[M] J. Milnor, Dynamics in one complex variable, Introductory lectures, , Preprint SUNY Stony Brook, 1990/5. [Nev] R. Nevanlinna Analytic functions, Springer 1970

[New] M. Newman, Elements Univ. Press, 1939

of the

Topology

of Plane

Sets, Cambridge
Math.

[P1] Ch. Pommerenke, On Ergodic Properties Ann. 256 (1981), 43-50.

of Inner Functions,

[P2] Ch. Pommerenke, Boundary Behavior of Conformal Maps, Springer Verlag, New York, Berlin and Heiderberg, 1992.
[S] G. M. Stallard, The iteration of entire functions Proc. Camb. Phil. Soc. 114 (1993), 43-55.

ofsmall growth, Math.