1: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
2: %
3: % This is the AMS-LaTeX source file for the paper,
4: %
5: %
6: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
7: \documentclass[12pt]{amsart}
8: \usepackage[]{amsmath, amsthm, amsfonts, amssymb, epsfig}
9: %\usepackage[]{showkeys}
10: \textheight222mm
11: \textwidth144mm
12: \topmargin-12mm
13: \oddsidemargin7.5mm
14: \parskip4pt plus2pt minus2pt
15:
16: \newcommand\C{{\mathbb C}}
17: \newcommand\R{{\mathbb R}}
18: \newcommand\dee{\partial}
19: \renewcommand\O{\Omega}
20: \newcommand\Obar{\overline{\Omega}}
21: \newcommand\Ot{\widetilde\Omega}
22: \newcommand\Ohat{\widehat\Omega}
23: \newcommand\ol{\overline}
24:
25: \numberwithin{equation}{section}
26:
27: \begin{document}
28:
29: \title[Two-connected domains]
30: {A Riemann mapping theorem for two-connected domains in the plane}
31: \author[Bell, Deger, Tegtmeyer]
32: {Steven R. Bell, Ersin Deger, \& Thomas Tegtmeyer}
33:
34: \address[]{Mathematics Department, Purdue University, West Lafayette,
35: IN 47907}
36: \email{bell@math.purdue.edu}
37:
38: \address[]{Mathematics Department, University of Cincinnati, Cincinnati,
39: OH 45221}
40: \email{ersin.deger@uc.edu}
41:
42: \address[]{Mathematics Department, Truman State University,
43: Kirksville, MO 63501}
44: \email{tegt@truman.edu}
45:
46: \subjclass{30C35}
47: \keywords{Ahlfors map, Bergman kernel, Szeg\H o kernel}
48:
49: \begin{abstract}
50: We show how to express a conformal map $\Phi$ of a general two connected
51: domain in the plane such that neither boundary component is a point to a
52: representative domain of the form ${\mathcal A}_r=\{z: |z+1/z|<2r\}$, where
53: $r>1$ is a constant. The domain ${\mathcal A}_r$ has the virtue of having
54: an explicit algebraic Bergman kernel function, and we shall explain why
55: it is the best analogue of the unit disc in the two connected setting.
56: The map $\Phi$ will be given as a simple and explicit algebraic function
57: of an Ahlfors map of the domain associated to a specially chosen point.
58: It will follow that the conformal map $\Phi$ can be found by solving
59: the same extremal problem that determines a Riemann map in the simply
60: connected case.
61: \end{abstract}
62:
63: \maketitle
64:
65: \theoremstyle{plain}
66:
67: \newtheorem {thm}{Theorem}[section]
68: \newtheorem {lem}[thm]{Lemma}
69:
70: \hyphenation{bi-hol-o-mor-phic}
71: \hyphenation{hol-o-mor-phic}
72:
73: \section{Main results}
74: \label{sec1}
75: The Riemann map $f_a$ associated to a point $a$ in a simply connected
76: domain $\O\ne\C$ in the complex plane maps $\O$ one-to-one onto the
77: unit disc $\{z: |z|<1\}$ and is the solution to an extremal problem: among
78: all holomorphic mappings of $\O$ into the unit disc, $f_a$ is the
79: map such that $f_a'(a)$ is real and as large as possible. The
80: objects of potential theory associated to the unit disc are
81: particularly simple. The Bergman and Szeg\H o kernel functions
82: are explicit and simple rational functions on the disc, for example.
83: The Poisson kernel is also very simple. Consequently, the Bergman,
84: Poisson, and Szeg\H o kernels associated to $\O$ can be expressed very
85: simply in terms of a Riemann map.
86:
87: It was noted in \cite{B4} that any two-connected domain $\O$ in the
88: plane such that neither boundary component is a point can be mapped
89: conformally via a mapping $\Phi$ to a representative domain of the form
90: $${\mathcal A}_r := \{z: |z+1/z|<2r\},$$
91: where $r$ is a constant greater than one. See \cite{J-T} for a nice
92: proof of this fact which also proves a much more general theorem.
93: See Figure~1 for a picture of ${\mathcal A}_r$. The dotted circle
94: in the figure is the unit circle. Notice that $1/z$ is an automorphism
95: of ${\mathcal A}_r$ that fixes the unit circle.
96:
97: In this paper, we show that the conformal map $\Phi:\Omega\to{\mathcal A}_r$
98: can be expressed simply in terms of an Ahlfors map of $\O$. The Bergman
99: kernel associated to a representative domain ${\mathcal A}_r$ is an
100: algebraic function (see \cite{B4,B5}). In fact, the Bergman kernel has
101: recently been written down explicitly in \cite{D}.
102: The Szeg\H o kernel is also algebraic and the Poisson kernel can
103: be expressed in terms of rather simple functions (see \cite{B6}). It is
104: proved in \cite{B3}
105: that neither the Bergman nor Szeg\H o kernel can be rational in a
106: two-connected domain. Thus, the representative domain ${\mathcal A}_r$
107: can be thought of as perhaps the best analogue of the unit
108: disc in the two-connected setting. We shall show that the conformal
109: map $\Phi$ can be determined by solving {\it the same\/} extremal problem
110: as the one that determines a Riemann map in the simply connected case, but
111: we must be careful to choose the base point properly. In fact, we
112: show how to find a point $a$ in $\O$ so that $\Phi$ is given by
113: $cf_a+\sqrt{c^2f_a^2-1}$ where $f_a$ is the Ahlfors mapping associated to
114: $a$, and $c$ is a complex constant that we will determine. We will
115: explain along the way how $cf_a+\sqrt{c^2f_a^2-1}$ can be understood to
116: be a single valued holomorphic function on $\O$. The Ahlfors
117: map $f_a$ is the solution to the extremal problem: among all holomorphic
118: mappings of $\O$ into the unit disc, $f_a$ is the map such that $f_a'(a)$
119: is real and as large as possible. See \cite[p.~49]{B2} for the basic
120: properties of the Ahlfors map.
121:
122: In order to choose $a$, we will inadvertently discover a new way to
123: compute the modulus of the domain $\O$.
124:
125: \begin{center}
126: \epsfig{file=peanut2.eps, height=3in}
127: \end{center}
128:
129: \begin{center}
130: Figure~1. The domain ${\mathcal A}_r$ with $r=1.05$.\\
131: \end{center}
132:
133: Let $a_\rho$ denote the annulus $\{z:\frac{1}{\rho}<|z|<\rho\}$ when $\rho>1$, and
134: let
135: $$J(z):=\frac{1}{2}\left(z+\frac{1}{z}\right)$$
136: denote the Joukowsky map. Note that $\frac{1}{r}J(z)$ is a proper holomorphic map of
137: ${\mathcal A}_r$ onto the unit disc which is a $2$-to-one (counting multiplicities)
138: branched covering map. Several authors have found explicit biholomorphic
139: maps between $a_\rho$ and ${\mathcal A}_r$ and have determined the
140: relationship between $\rho$ and $r$ (see \cite{J-T,JOT,C-M,D}). Since we shall
141: need such a map, and since we can construct one in a minimum of space here using
142: the method from \cite{D}, we shall include the construction for completeness.
143: The biholomorphic mapping $\Psi: a_\rho \to {\mathcal A}_r$ will be an explicit
144: map involving the Jacobi {\it sn\/} function. There is a standard conformal map
145: of the annulus $\{z: \frac{1}{\rho}<|z|<1\}$ onto the unit disc minus a slit from
146: $-L$ to $+L$ along the real axis. See Nehari \cite[p.~293]{N}) for a construction
147: where the relationship between $L$ and $\rho$ is given explicitly. The mapping is
148: given by a constant times
149: $$\text{\it sn}\left(K+\frac{2iK}{\pi}\log\rho z\, ;\, \rho^{-4}\right),$$
150: where $K$ is a constant that can be determined explicitly. The map takes the
151: unit circle to the unit circle and the circle of radius~$1/\rho$ to the slit from
152: $-L$ to $L$. By composing with an inversion of the annulus, we get a conformal
153: map that maps the annulus $\{z: \frac{1}{\rho}<|z|<1\}$ one-to-one onto the unit
154: disc minus a slit from $-L$ to $L$ that maps the unit circle to the slit from
155: $-L$ to $L$ and maps the circle of radius~$1/\rho$ to the unit circle.
156: The Joukowsky map $J(z)$ is a biholomorphic mapping of $\{z\in
157: {\mathcal A}_r : |z|>1\}$ onto the disc of radius $r$ minus a slit
158: from $-1$ to $+1$. It maps the outer boundary of ${\mathcal A}_r$ to the unit
159: circle and the unit circle two-to-one onto the slit. Thus, by scaling properly
160: and then composing, we can write down a biholomorphic mapping from
161: the annulus $\{z: \frac{1}{\rho}<|z|<1\}$ onto $\{z\in{\mathcal A}_r : |z|>1\}$
162: which maps the unit circle to the unit circle. We may
163: now extend this map by reflection to a biholomorphic map from $a_\rho$ onto
164: ${\mathcal A}_r$. Finally, we may compose with the automorphism $1/z$ and a
165: rotation, if necessary, to obtain a conformal mapping
166: $\Psi: a_\rho\to{\mathcal A}_r$ such that the outer boundary of $a_\rho$ maps
167: to the outer boundary of ${\mathcal A}_r$ and such that $\Psi(i)=i$. Note
168: that $\Psi$ is uniquely determined by these conditions. We shall need to know
169: that $\Psi(-i)=-i$. Since $-1/\Psi(-1/z)$ satisfies the same properties,
170: namely that it is a conformal map that takes the outer boundary to the outer
171: boundary and $\Psi(i)=i$, it must be that
172: $\Psi(z)\equiv -1/\Psi(-1/z)$. From this, we deduce that
173: $\Psi(-i)\equiv -1/\Psi(-i)$, and hence that $\Psi(-i)$ is a square root
174: of $-1$. Because $\Psi$ is one-to-one, it cannot be that $\Psi(-i)$ is
175: $i$, and so $\Psi(-i)=-i$.
176:
177: It is well known that any two-connected domain $\O$ in the plane such that neither
178: boundary component is a point can be mapped biholomorphically via a map $F$ to a unique
179: annulus of the form $a_\rho$. Let $C_1$ denote the unit circle. We shall call
180: $\rho^2$ the modulus of $\O$ and we shall call the set $F^{-1}(C_1)$ the {\it median\/}
181: of the two-connected domain. The median is the unique connected one-dimensional set
182: that is left invariant by every automorphism of the two-connected domain. The
183: biholomorphic map $\Psi$ of $a_\rho$ onto ${\mathcal A}_r$ preserves the unit circle.
184: Hence, the median of ${\mathcal A}_r$ is also the unit circle.
185:
186: Our main results will follow from two key lemmas, which we now list.
187:
188: \begin{lem}
189: \label{lem1}
190: Suppose that $\O$ is a two-connected domain in the plane such that neither
191: boundary component is a point. The Ahlfors map $f_a$ associated to any point $a$
192: in $\O$ is such that it has two distinct and simple branch points on the median
193: of $\O$.
194: \end{lem}
195:
196: Lemma~\ref{lem1} was proved by McCullough and Mair in \cite{M-M} and
197: independently and by other means in \cite{Te}.
198:
199: \begin{lem}
200: \label{lem2}
201: The Ahlfors map associated to the point $i$ in
202: ${\mathcal A}_r$ is given by $\frac{1}{r}J(z)$.
203: \end{lem}
204:
205: Lemma~\ref{lem2} was proved in \cite{D}. For completeness, and because it
206: will not take much space, we will give alternate proofs of the lemmas in
207: the next section.
208:
209: Suppose that $\O$ is a
210: two-connected domain in the plane such that neither boundary component
211: is a point. We now describe the procedure to determine the mapping $\Phi$,
212: assuming the truth of the lemmas. Pick any point $P$ in $\O$. Lemma~\ref{lem1}
213: guarantees the existence of a point $a$ on the median of $\O$ such that $f_P'(a)=0$.
214: (Note that the Ahlfors map is eminently numerically computable if the boundary
215: of $\O$ is sufficiently smooth (see \cite{K-S,K-T,Tr,B1}), and the zeroes of a
216: function with only two simple zeroes can easily be located numerically as well.)
217: Now the Ahlfors map $f_a$ associated to the point $a$ is such that there are
218: two distinct points $p_1$ and $p_2$ on the median of $\O$ such that $f_a'(p_i)=0$
219: for $i=1,2$. The points $p_i$ are also different from $a$ because $f_a'(a)\ne0$.
220: We now claim that there
221: exists an $r>1$ and a biholomorphic mapping from $\O$ to ${\mathcal A}_r$ such
222: that $\Phi(a)=i$. We may also arrange for the outer boundary of $\O$ to get mapped
223: to the outer boundary of ${\mathcal A}_r$. To see this, we use the
224: biholomorphic mapping $F$ of $\O$ onto an annulus $a_\rho$.
225: By composing with $1/z$, if necessary, we may map the outer boundary of $\O$ to
226: the outer boundary of $a_\rho$. Since $a$ is on the median, $F(a)$ is on the
227: unit circle, and we may compose with a rotation so that $F(a)=i$. Now we
228: may compose with the biholomorphic map $\Psi$ constructed above of
229: $a_\rho$ onto ${\mathcal A}_r$, noting that $i$ is fixed under $\Psi$ and that
230: outer boundaries are preserved, to obtain $\Phi$. (At the moment, we can
231: only say that the number $r$ exists. We will explain how to find it
232: momentarily.) Lemma~\ref{lem2} states that $\frac{1}{r}J(z)$ is the Ahlfors
233: map of ${\mathcal A}_r$ associated to the point $i$. The extremal property of
234: Ahlfors maps makes it easy to see that there is a unimodular complex number
235: $\lambda$ such that
236: \begin{equation}
237: \label{eqn1}
238: \frac{1}{r}J(\Phi(z))=\lambda f_a(z).
239: \end{equation}
240: Since the branch points of $J(z)$ occur at $\pm 1$, this last identity reveals
241: that $\Phi(p_i)=\pm 1$ for $i=1,2$. We may assume that $\Phi(p_1)=-1$ and
242: $\Phi(p_2)=1$. Since $J(-1)=-1$ and $J(1)=1$, equation~(\ref{eqn1}) now gives
243: $-\frac{1}{r}=\lambda f_a(p_1)$ and $\frac{1}{r}=\lambda f_a(p_2)$.
244: Either one of these two equations determines both $\lambda$ and $r$.
245: We can now solve equation~(\ref{eqn1}) locally for $\Phi$ to obtain
246: $$\Phi(z)=J^{-1}(c f_a(z)),$$
247: where $c=r\lambda$ and, of course, $J^{-1}(w)=w+\sqrt{w^2-1}$.
248: This formula extends by analytic continuation from any point in $\O$
249: that is not a branch point of $f_a$. The apparent singularities at
250: the branch points are removable because the construction aligned the
251: two algebraic singularities of $J^{-1}$ to occur at the two images
252: of the branch points under $cf_a$. We shall explain how to compute
253: $\Phi$ from its boundary values in \S\ref{sec3}.
254: The modulus $\rho^2$ of $\O$ can now be gotten from $r$ via any of
255: the formulas relating $\rho$ and $r$ proved in \cite{J-T,JOT,C-M,D}.
256:
257: In the last section of this paper, we explore how the mapping $\Phi$
258: can be used to study the Bergman kernel of two-connected domains.
259:
260: \section{Proofs of the Lemmas}
261: \label{sec2}
262:
263: To prove Lemmas~\ref{lem1} and~\ref{lem2}, we will make use of the simple
264: fact that if $G:\O_1\rightarrow\O_2$ is a biholomorphism between bounded
265: domains, the Ahlfors maps transform according to the following formula. Let
266: $f_j(z;w)$ denote the Ahlfors map $f_w$ associated to the point $w$ in $\O_j$.
267: The extremal property of the Ahlfors maps can be exploited to
268: show that
269: \begin{equation}
270: \label{eqn2}
271: f_1(z;w)=\lambda f_2\left(G(z);G(w)\right),
272: \end{equation}
273: where $\lambda$ is the unimodular constant that makes the derivative of
274: $f_1(z;w)$ in the $z$ variable at $z=w$ real and positive.
275:
276: \noindent
277: {\bf Proof of Lemma~\ref{lem1}.}
278: The biholomorphic map $F$ from $\O$ to $a_\rho$ together with
279: equation~(\ref{eqn2}) reveals that it is sufficient to prove Lemma~\ref{lem1}
280: for an annulus $a_{\rho}$ with base point $a>0$. The function $h(z)=e^{iz}$
281: maps the horizontal strip $H=\{z\,:\,-\log \rho<{\rm Im}\, z<\log \rho\}$
282: onto $a_{\rho}$, takes the real line onto $C_1,$ and is $2\pi$-periodic.
283: Consider $g(z)=f_a(h(z))$. Notice that $g$ is $2\pi$-periodic, maps $H$
284: onto the unit disk, and maps the boundary of $H$ onto the unit circle.
285: It can be extended via a sequence of reflections to a function on $\C$.
286: Because of these reflections, $g$ has a second period of $(4\log\rho)i,$
287: making $g$ an elliptic function. Because $f_a$ is two-to-one, $g$ is
288: a second order elliptic function.
289:
290: Any second order elliptic function $\phi$ satisfies the equation
291: $\phi(w-z)=\phi(z),$ where $w$ is the sum of two noncongruent zeroes of
292: $\phi$ (see \cite{H}). From \cite{T-T}, we know that if $a$ is real,
293: $f_a$ has zeroes at $a$ and $-1/a$. (We will prove this fact below to make
294: this paper self contained.) It now follows that $g$ will have zeroes at
295: $-i\log a$ and $i\log a+\pi.$ Thus, we can choose $w=\pi.$ The mapping
296: $z\mapsto \pi-z$ has $\pi/2$ as a fixed point, while the point $3\pi/2$ gets
297: mapped to the congruent point $-\pi/2$. These two points will be branch
298: points of $g,$ and $i$ and $-i$ will be the branch points of $f_a$,
299: lying on the median of $a_{\rho}.$
300:
301: \noindent
302: {\bf Proof of Lemma~\ref{lem2}.}
303: The biholomorphism $\Psi$ (see \S\ref{sec1}) from $a_{\rho}$ onto
304: $\mathcal A_r$ mapping $i$ to $i$ maps $-i$ to $-i$. Let $f_i$
305: denote the Ahlfors map of $a_{\rho}$ with base point $i$. From
306: \cite{T-T}, we know that $f_i$ will have $i$ and $-i$ as its zeroes.
307: (We include a proof of this fact in the next paragraph.)
308: Equation~(\ref{eqn2}) tells us that the Ahlfors map of ${\mathcal A}_r$
309: with base point $i$ will have $i$ and $-i$ as its zeroes. Because
310: this Ahlfors map has the same zeroes as $\frac{1}{r}J(z),$ and both
311: functions map the boundary of $\mathcal A_r$ to the unit circle, they
312: are the same up to multiplication by a unimodular constant. Since
313: $J'(i)=1>0$, this constant must be $1$. Lemma~\ref{lem2} is proved.
314:
315: We shall now flesh out the proofs of the lemmas by giving a shorter
316: and simpler proof than the one given in \cite{T-T} that the
317: Ahlfors map $f_a$ associated to a point $a$ in $a_\rho$ has simple
318: zeroes at $a$ and $-1/\bar{a}$. Thus $f_a$ has simple zeroes at
319: $a$ and $-1/a$ if $a$ is real, and $f_i$ has simple zeroes at $i$
320: and $-i$. The Ahlfors map $f_a$ always has a simple zero at $a$.
321: Since the Ahlfors map is two-to-one (counting multiplicities), $f_a$
322: must have exactly one other simple zero distinct from $a$.
323: The Ahlfors map is given via
324: \begin{equation}
325: \label{eqnSL}
326: f_a(z)=S(z,a)/L(z,a)
327: \end{equation}
328: where $S(z,a)$ is the Szeg\H o kernel and $L(z,a)$ is the
329: Garabedian kernel. For the basic properties of the Szeg\H o and
330: Garabedian kernels, see \cite[p.~49]{B2}. We note here that
331: $S(z,a)=\overline{S(a,z)}$ and that $L(z,a)\ne0$ if $a\in\O$ and
332: $z\in\Obar$ with $z\ne a$. Also, $L(z,a)$ has a single simple pole
333: at $a$ in the $z$ variable, and $L(z,a)=-L(a,z)$. It follows from
334: equation~(\ref{eqnSL}) that the other zero of $f_a$ is the one and
335: only zero of $S(z,a)$ in the $z$ variable. Let $Z(a)$ denote
336: this zero. It is proved in \cite{B4} that $Z(a)$
337: is a proper antiholomorphic self-correspondence of the domain. In
338: the two-connected case that we are in now, it is rather easy to see that
339: $Z(a)$ is an antiholomorphic function of $a$ that maps $a_\rho$ one-to-one
340: onto itself. To see this, let $S'(z,a)$ denote the derivative of $S(z,a)$
341: in the $z$ variable and notice that the residue theorem yields
342: $$Z(a)=\frac{1}{2\pi i}\int_{z\in b\O}\frac{ z S'(z,a)}{S(z,a)}\ dz.$$
343: (See \cite{B2} for proofs that $S(z,a)$ is smooth up to the boundary
344: and non-vanishing on the boundary in the $z$ variable when $a$ is held
345: fixed in $a_\rho$.) It can be read off from this last formula that
346: $Z(a)$ is antiholomorphic in $a$. Now hold $z$ fixed in $a_\rho$ and
347: let $a$ tend to a boundary point $a_b$. Equation~(\ref{eqnSL})
348: shows that $f_a(z)$ tends to $S(z,a_b)/L(z,a_b)$. But
349: $S(z,a_b)=-\frac{1}{i}L(z,a_b)T(a_b)$ where $T(a_b)$ is the complex
350: unit tangent vector at $a_b$ (see \cite[p.~107]{B2}), and
351: so $f_a(z)$ tends to $iT(a_b)$, a unimodular constant. This shows
352: that the zeroes of $f_a(z)$ must tend to the boundary as $a$ tends to
353: the boundary. Consequently, $Z(a)$ is a proper antiholomorphic self mapping
354: of the annulus. The only such maps are one-to-one and onto (see \cite{M-R}).
355: We shall say that $Z(a)$ is an antiholomorphic automorphism of
356: $a_{\rho}$. Note that $Z(a)$ must satisfy $Z(Z(a))=a$. The only such
357: antiholomorphic automorphisms are $Z(a)=\lambda\bar{a}$, where $\lambda$
358: is a unimodular constant, and $Z(a)=1/\bar{a},$ and $Z(a)=-1/\bar{a}$.
359: Since $f_a$ does not have a double zero at $a$, $Z(a)$ cannot have a fixed
360: point. Only $Z(a)=-1/\bar{a}$ is without fixed points in $a_{\rho},$ so
361: $f_a$ has zeroes at $a$ and $-1/\bar{a}$. The proof is complete.
362:
363: \section{How to compute $\Phi$ and the median of $\O$.}
364: \label{sec3}
365: Suppose that $\O$ is a bounded two connected domain bounded by two
366: non-intersecting $C^\infty$ Jordan curves. The Kerzman-Stein-Trummer
367: method \cite{K-S,K-T} can be used to compute the boundary values of
368: Ahlfors mappings associated to $\O$. (See \cite{B1} and \cite{B2},
369: Chapter~26, for a description of how to do this in the multiply
370: connected setting.) Thus, we may pick a point $P$ in $\O$ and
371: compute $f_P$. We shall explain how to compute the two branch
372: points of $f_P$ on the median of $\O$ below. Hence, we may find
373: a point $a$ on the median of $\O$. Let $f$ denote the Ahlfors
374: mapping associated to the point $a$ and let $p_1$
375: and $p_2$ denote the two simple zeroes of $f'$ on the median of $\O$.
376: We shall now show how to compute the points $p_1$ and $p_2$
377: and the function $\Phi$ using only the boundary values of the Ahlfors
378: map $f$. The residue theorem yields that
379: $$p_1^n+p_2^n=\frac{1}{2\pi i}\int_{b\O} \frac{f''(z)z^n}{f'(z)}\ dz.$$
380: The two points can be determined from the values of
381: $p_1^n+p_2^n$ for $n$ equal to one and two. Indeed the coefficients
382: $A$ and $B$ of the polynomial $z^2-Az+B=(z-p_1)(z-p_2)$ are given
383: as $A=p_1+p_2$ and $B=\frac{1}{2}[(p_1+p_2)^2-(p_1^2+p_2^2)]$, and
384: so $p_1$ and $p_2$ can be found via the quadratic formula. Therefore,
385: the point $a$ itself and the branch points of $f_a$ can readily be found
386: numerically.
387:
388: We may calculate $c=r\lambda=-1/f(p1)$ by the procedure
389: described in \S\ref{sec1}. There are two holomorphic branches of
390: $J^{-1}(w)=w+\sqrt{w^2-1}$ near the boundary of the disc of radius~$r$
391: corresponding to the two choices of the square root. One branch $j_o$
392: maps the circle of radius~$r$ one-to-one onto the outer boundary of ${\mathcal A}_r$
393: and the other branch $j_i$ maps the circle of radius~$r$ one-to-one onto the inner
394: boundary of ${\mathcal A}_r$. The mapping $\Phi$ that maps $\O$ one-to-one
395: onto ${\mathcal A}_r$ can be determined from its boundary values on $\O$.
396: The boundary values of $\Phi$ on the outer boundary of $\O$ can be taken to
397: be $j_o(cf(z))$ and the boundary values on the inner boundary are then
398: $j_i(cf(z))$. The extension of $\Phi$ to $\O$ can be gotten from the
399: boundary values of $\Phi$ via the Cauchy integral formula.
400:
401: We now turn to a method to determine the median of $\O$. Equation~(\ref{eqn1})
402: shows that $J(\Phi(z))= c f(z)$ where $c=\lambda r$ and $f$ is shorthand for
403: $f_a$. This shows that $f$ maps the median of $\O$ to the line segment
404: ${\mathcal L}$ from $-1/c$ to $1/c$. Hence, the median of $\O$ is given
405: by $f^{-1}({\mathcal L})$. We already know the points $p_1$, $p_2$, and $a$
406: on the median. To compute other points $w$ in this set such that $f(w)=\tau$
407: for some $\tau$ in ${\mathcal L}$ where $\tau$ is not one of the endpoints
408: of ${\mathcal L}$, we may again use the residue theorem to obtain
409: $$w_1^n +w_2^n=\frac{1}{2\pi i}\int_{z\in b\O}\frac{f'(z)z^n}{f(z)-\tau}\ dz,$$
410: where $w_1$ and $w_2$ are the two points in $f^{-1}(\tau)$. Finally,
411: we may use the Newton indentity for $n=2$ and the quadratic formula
412: as we did above to determine $p_1$ and $p_2$ to get $w_1$ and $w_2$.
413: Note that $p_1$ and $p_2$ correspond to the inverse images of the two
414: endpoints of ${\mathcal L}$, and so we may generate the rest of the median
415: by letting $\tau$ move between the endpoints.
416:
417: \section{Other consequences.}
418: We describe in this section how the Bergman kernel of a two-connected domain
419: can be expressed rather concretely in terms of an Ahlfors map.
420: It was proved in \cite{D} that the Bergman kernel associated to ${\mathcal A}_r$
421: is given by
422: $$K_r(z,w)=C_1\frac{2k^2S(z,\bar w)+kC(z,\bar w)D(z,\bar w)+ C_2}
423: {z\bar w\sqrt{1-k^2J(z)^2}\sqrt{1-k^2J(\bar w)^2}},$$
424: where $k=1/r^2$, and $C_1$ and $C_2$ are constants that only depend on $r$, and
425: $$S(z,w)=-\left(\frac
426: {J(z)\frac{w^2-1}{2w}\sqrt{1-k^2J(z)^2}+
427: J(w)\frac{z^2-1}{2z}\sqrt{1-k^2J(w)^2}}
428: {1-k^2J(z)^2J(w)^2}\right)^2,$$
429: and
430: $$C(z,w)=
431: \frac
432: {-\frac{z^2-1}{2z}\frac{w^2-1}{2w}-J(z)J(w)\sqrt{1-k^2J(z)^2}\sqrt{1-k^2J(w)^2}}
433: {1-k^2J(z)^2J(w)^2},$$
434: and
435: $$D(z,w)=
436: \frac
437: {\sqrt{1-k^2J(z)^2}\sqrt{1-k^2J(w)^2}+k^2J(z)J(w)\frac{z^2-1}{2z}\frac{w^2-1}{2w}}
438: {1-k^2J(z)^2J(w)^2}.$$
439: We showed that the biholomorphic map $\Phi$ that we constructed above from a
440: two-connected domain to its representative domain ${\mathcal A}_r$ is such that
441: $\frac{1}{r}J(\Phi(z))=\lambda f_a(z)$, where $f_a$ is an Ahlfors map associated
442: to $\O$ for a point $a$ on the median of $\O$ and $\lambda$ is a unimodular
443: constant. Differentiate this identity to see that
444: $$\Phi'(z)=2c f_a'(z)/(1-\Phi(z)^{-2})=
445: 2c f_a'(z)/(1-J^{-1}(cf_a(z))^{-2}),$$
446: where $J^{-1}(w)=w+\sqrt{w^2-1}$.
447: The Bergman kernel associated to $\O$ is given by
448: $$\Phi'(z)K_r(\Phi(z),\Phi(w))\overline{\Phi'(w)}.$$
449: We may read off from this and the formulas above that the Bergman kernel
450: associated to $\O$ is $f_a'(z)\,\overline{f_a'(w)}$ times a rational function
451: of $f_a(z)$, and $\overline{f_a(w)}$, and simple algebraic functions of
452: $f_a$ of the form
453: $\sqrt{A f_a(z)^2-1}$ and the conjugate
454: of $\sqrt{Af_a(w)^2-1}$, where $A$ is equal to $c$, $\lambda$, or one.
455: The degree of the rational function does not depend on the modulus of
456: the domain, and the coefficients of the rational function depend in a
457: straightforward way on the modulus.
458:
459: It was proved in \cite{B5} that the Bergman kernel associated to a
460: two connected domain such that neither boundary component is a point
461: can be expressed in terms of two Ahlfors maps $f_a$ and $f_b$ as
462: $$f_a'(z)R(z,w)\overline{f_a'(w)}$$
463: where $R(z,w)$ is a rational function of $f_a(z)$, $f_b(z)$,
464: $\overline{f_a(w)}$, and $\overline{f_b(w)}$. The
465: rational function has been proved to exist, but it has never been
466: given explicitly. We are now in a position to write down this rational
467: function for the domain ${\mathcal A}_r$ using the points $a=i$ and
468: $b=1$. It was proved in \cite{D} that the Ahlfors map associated
469: to the point $1$ of ${\mathcal A}_r$ is
470: $$g(z)=\frac{1}{2r}\frac{z-\frac{1}{z}}{\sqrt{1-k^2J(z)^2}}$$
471: where $k=1/r^2$. The square root in the formula can be taken to be
472: the principal branch because the modulus of $kJ(z)$ is less than one
473: in ${\mathcal A}_r$. We shall prove this result here because all
474: the ingredients are on the table. It is a pleasant exercise to check
475: that $g$ is a two-to-one branched covering map of ${\mathcal A}_r$ onto
476: the unit disc. Notice that $g$ has zeroes at $1$ and $-1$.
477: We proved above that the Ahlfors map $f_1$ has zeroes
478: at $1$ and $-1$. Hence $f_1/g$ is a unimodular constant. Checking
479: derivatives at the point $1$ yields that the unimodular constant is
480: $1$. So $g$ is equal to the Ahlfors map $f_1$. Recall that the
481: Ahlfors map $f_i$ associated to $i$ is just $\frac{1}{r}J(z)$.
482:
483: To determine the rational function, we make note that the terms
484: $\frac{z^2-1}{2z}$ can be multiplied by unity in the form of
485: $$\frac{\sqrt{1-k^2J(z)^2}}{\sqrt{1-k^2J(z)^2}}$$
486: to see that
487: $$\frac{z^2-1}{2z}= r\frac{1}{2r}\left(z-\frac{1}{z}\right)=
488: rf_1(z)\sqrt{1-k^2J(z)^2}
489: =rf_1(z)\sqrt{1-r^2f_i(z)^2}.$$
490: The terms in the denominator in the expression for $K(z,w)$ can
491: be manipulated by multiplying by unity in the form of $f_i'/f_i'$,
492: noting that $f_i'(z)=\frac{1}{r}(1-\frac{1}{z^2})$,
493: to see that
494: $$z\sqrt{1-k^2J(z)}=\frac{2}{f_i'(z)} f_1(z)(1-r^2f_i(z)^2).$$
495: When these procedures are carried out in each of the expressions
496: that comprise the Bergman kernel, and when all the terms are
497: combined, the square roots all become squared and we obtain
498: $$K(z,w)=f_i'(z)
499: Q(f_i(z),f_1(z),\overline{f_i(w)},\overline{f_1(w)})
500: \overline{f_i'(w)},$$
501: where $Q(z_1,z_2,w_1,w_2)$ is a rational function of four
502: variables given by
503: $$C_1\left(\frac{\sigma(z_1,z_2,w_1,w_2)+\delta(z_1,z_2,w_1,w_2) +C_2}
504: {q(z_1,z_2,w_1,w_2)}\right),$$
505: where
506: $$\sigma(z_1,z_2,w_1,w_2)=-2\left(\frac{z_1w_2(1-r^2w_1^2)+w_1z_2(1-r^2z_1^2)}
507: {1-z_1^2w_1^2}
508: \right)^2$$
509: and
510: $$\delta(z_1,z_2,w_1,w_2)=-\left(\frac{(z_2w_2+z_1w_1)(1-r^2z_1^2)(1-r^2w_1^2)(1+z_1w_1z_2w_2) + C_2}
511: {(1-z_1^2w_1^2)^2}
512: \right),$$
513: and
514: $$q(z_1,z_2,w_1,w_2)=z_2w_2(1-r^2z_1^2)(1-r^2w_1^2).$$
515: That is more than twice as complicated as any formula on the
516: one-connected unit disc, but it could be worse. A fascinating feature
517: that can be read off is that the coefficients depend very simply on $r$
518: and the degree of the rational function does not depend on $r$.
519:
520: For the general two connected domain $\O$, we use the mapping $\Phi$
521: and the transformation formula for the Bergman kernels to read off that
522: $$K(z,w)=f_a'(z)R(z,w)\overline{f_a'(w)},$$
523: where $R(z,w)$ is a rational function of $f_a(z)$, $f_b(z)$,
524: $\overline{f_a(w)}$, and $\overline{f_b(w)}$. We take
525: $a$ to be the point we chose in the construction of $\Phi$ and
526: $b$ to be equal to $p_2$. We obtain a formula for the rational function
527: and we note that the degree does not depend on $r$.
528:
529: It will be interesting to investigate if similar properties hold for
530: domains of higher connectivity using techniques of Crowdy and Marshall
531: \cite{C-M}.
532:
533: \begin{thebibliography}{JOT}
534:
535: \bibitem[B1]{B1}
536: Bell, Steven~ R., {\em Numerical computation of the Ahlfors map of a multiply
537: connected planar domain}. J. Math. Anal. Appl. {\bf 120} (1986), 211--217.
538:
539: \bibitem[B2]{B2}
540: \rule{3em}{.5pt}\thinspace, {\em The Cauchy transform, potential theory, and
541: conformal mapping}, CRC Press, Boca Raton, 1992.
542:
543: \bibitem[B3]{B3}
544: \rule{3em}{.5pt}\thinspace, {\em Complexity of the classical kernel functions of
545: potential theory}. Indiana Univ. Math. J. {\bf 44} (1995), 1337--1369.
546:
547: \bibitem[B4]{B4}
548: \rule{3em}{.5pt}\thinspace, {\em Finitely generated function fields and
549: complexity in potential theory in the plane}, Duke Math. J. {\bf 98 }
550: (1999), 187--207.
551:
552: \bibitem[B5]{B5}
553: \rule{3em}{.5pt}\thinspace,
554: {\em Complexity in complex analysis}, Advances in Math. {\bf 172}
555: (2002), 15--52.
556:
557: \bibitem[B6]{B6}
558: \rule{3em}{.5pt}\thinspace,
559: {\em The Greens function and the Ahlfors map}, to appear.
560:
561: \bibitem[C]{C}
562: D.~Crowdy, {\em Conformal mappings from annuli to canonical doubly
563: connected Bell representations,} to appear.
564:
565: \bibitem[C-M]{C-M}
566: D.~Crowdy and M.~Marshall, {\em Conformal mappings between canonical multiply
567: connected planar domains,} Computational Methods and Function Theory {\bf 6}
568: (2006), 59--76.
569:
570: \bibitem[D]{D}
571: Ersin Deger, {\em A biholomorphism from the Bell representative domain onto
572: an annulus and kernel functions}, Purdue Univ. PhD thesis, 2007.
573:
574: \bibitem[H]{H}
575: E.~Hille, {\em Analytic Function Theory}, vol.~2, Blaisdell Publ. Co.,
576: Waltham, MA, 1962.
577:
578: \bibitem[J-T]{J-T}
579: M.~Jeong and M.~Taniguchi, {\em Bell representations of finitely
580: connected planar domains}, PAMS {\bf 131} (2003), 2325--2328.
581:
582: \bibitem[JOT]{JOT}
583: M.~Jeong, J.-W.~Oh, and M.~Taniguchi, {\em Equivalence problem
584: for annuli and Bell representations in the plane,} J. Math. Anal.
585: Appl. {\bf 325} (2007), 1295-1305.
586:
587: \bibitem[K-S]{K-S}
588: N.~Kerzman and E.~M.~Stein, {\em
589: The Cauchy kernel, the Szeg\H o kernel, and the Riemann mapping
590: function}, Math. Annalen {\bf 236} (1978), 85--93.
591:
592: \bibitem[K-T]{K-T}
593: N.~Kerzman and M.~Trummer,
594: {\it Numerical conformal mapping via the Szeg\H o kernel},
595: in Numerical conformal mapping,
596: L. N. Trefethen, ed., North Holland,
597: Amsterdam, 1986.
598:
599: \bibitem[M-M]{M-M}
600: Mair, B. A. and S.~McCullough, {\em Invariance of extreme harmonic functions
601: on an annulus; applications to theta functions}. Houston J. Math. {\bf 20}
602: (1994), 453--473.
603:
604: \bibitem[M-R]{M-R}
605: Mueller, Carl and Walter Rudin, {\em Proper holomorphic self-maps of plane regions}.
606: Complex Variables {\bf 17} (1991), 113-121.
607:
608: \bibitem[N]{N}
609: Nehari, Z., {\em Conformal Mapping}, Dover, New York, 1952.
610:
611: \bibitem[Te]{Te}
612: Thomas Tegtmeyer, {\em The Ahlfors map and the Szeg\H o kernel in multiply
613: connected domains}, Purdue Univ. PhD thesis, 1998.
614:
615: \bibitem[T-T]{T-T}
616: Thomas Tegtmeyer and Anthony Thomas, {\em The Ahlfors map and the Szeg\H o kernel for
617: an annulus}, Rocky Mountain Math. J. {\bf 20} (1999), 709--723.
618:
619: \bibitem[Tr]{Tr}
620: M.~Trummer,
621: {\em An efficient implementation of a conformal mapping method based
622: on the Szeg\H o kernel}, SIAM J. of Numer. Anal. {\bf 23} (1986),
623: 853--872.
624:
625: \end{thebibliography}
626:
627: \end{document}
628: