1: \documentclass[11pt]{article}
2: \usepackage{epsfig}
3: \topmargin=-0.5cm \textwidth=16cm \oddsidemargin=-0.5cm
4: \textheight=22cm
5: \begin{document}
6:
7: \title{Discrete
8: $Z^{\gamma}$: embedded circle patterns with the combinatorics of
9: the square grid and
10: discrete Painlev\'e equations}
11:
12: \author{{\Large
13: Agafonov S.I. } \\
14: \\ Department of Mathematical Sciences \\
15: Loughborough University \\
16: Loughborough, Leicestershire LE11 3TU \\
17: United Kingdom \\
18: e-mail: {\tt
19: S.Agafonov@lboro.ac.uk} }
20: \date{}
21: \maketitle
22:
23: \unitlength=1mm
24:
25: \newtheorem{theorem}{Theorem}
26: \newtheorem{proposition}{Proposition}
27: \newtheorem{lemma}{Lemma}
28: \newtheorem{corollary}{Corollary}
29: \newtheorem{definition}{Definition}
30:
31: \pagestyle{plain}
32:
33: \begin{abstract}
34: A discrete analogue of the holomorphic map $z^{\gamma }$ is
35: studied. It is given by Schramm's circle pattern with the
36: combinatorics of the square grid. It is shown that the
37: corresponding
38: circle patterns are
39: embedded and described by special separatrix solutions of discrete
40: Painlev\'e equations. Global properties of these solutions, as
41: well as of the discrete $z^{\gamma }$, are established.
42: \end{abstract}
43:
44: \section{Introduction}
45:
46:
47: The theory of circle patterns is a fast developing field of
48: research on the border of complex analysis and discrete geometry.
49: Recent progress in this area has its origin in Thurston's idea
50:
51: \cite{T} about approximating the Riemann mapping by circle
52: packings. By now some aspects of the theory of circle patterns, as
53: discrete analogs of conformal mappings, are well understood, while
54: the others are still waiting to be clarified. Classical circle
55: packings comprised of disjoint open disks were later generalized
56: to circle patterns where the disks may overlap (see for example
57: \cite{H}). Different underlying combinatorics were considered.
58: Schramm introduced a class of circle patterns with the
59: combinatorics of the square grid \cite{Schramm};
60: hexagonal
61: circle patterns with constant multi-ratios were studied by
62: Bobenko, Hoffman and Suris in \cite{BHS}; a new rich class of
63: hexagonal patterns with constant intersection angles was
64: investigated in \cite{BH}.
65:
66: The convergence of discrete conformal maps represented by circle
67: packings was proven by Rodin and Sullivan \cite{RS}. For a
68: prescribed regular combinatorics this result was refined. He and
69: Schramm \cite{HS} showed that for hexagonal packings the
70: convergence is $C^{\infty}.$ The uniform convergence for circle
71: patterns with the combinatorics of the square grid and orthogonal
72: neighboring circles was established by Schramm \cite{Schramm}.
73:
74: Other facts underlining the striking analogy between circle
75: patterns and the classical theory are the uniformization theorem
76: concerning circle packing realizations of
77: cell complexes of a prescribed combinatorics \cite{BS}, discrete
78: maximum principle, Schwarz's lemma
79: and rigidity properties \cite{MR},\cite{H}, discrete Dirichlet
80: principle \cite{Schramm}.
81:
82: It turned out that an effective approach to the description and
83: the construction of
84: circle patterns with
85: overlapping circles is given by the theory of integrable systems
86: (see \cite{BPD},\cite{BHS},\cite{BH}). In particular, Schramm's
87: circle patterns studied in this paper are governed by a difference
88: equation which is the stationary Hirota equation (see
89: \cite{Schramm}). This equation is an example of an integrable
90: difference equation. It appeared first in a different branch of
91: mathematics -- the theory of integrable systems (see \cite{Z} for
92: a good survey).
93:
94: On the other hand, not very much is known about analogs of
95: standard holomorphic functions, although computer experiments give
96: evidence for their existence \cite{DS}. For circle packings with
97: the hexagonal combinatorics the only explicitly
98: described examples are Doyle spirals, which are discrete analogs
99: of exponential maps, \cite{Doy} and conformally symmetric
100: packings, which are analogs of a quotient of Airy functions
101: \cite{BHConf}. For patterns with overlapping circles more explicit
102: examples are known: discrete versions of ${\rm exp}(x)$,
103: ${\rm erf}(x)$ \cite{Schramm},
104: $z^{\gamma}$, $ {\rm log}(x)$ \cite{AB} are constructed for
105: patterns with underlying combinatorics of the square grid;
106: $z^{\gamma}$, ${\rm log}(x)$ are also described for hexagonal
107: patterns
108: with both multi-ratio \cite{BHS}
109: and constant angle \cite{BH} properties.
110:
111: Whereas computer experiments reveal a regular behavior of the
112: circle patterns corresponding to discrete
113: $z^{\gamma}$ and ${\rm log}(x)$, only the local property of
114: immersionness was proved for the Schramm's patterns \cite{AB}.
115: This
116: property turned out to be connected with special solutions of
117: discrete Painlev\'e II equations, thus giving geometrical
118: interpretation thereof.
119: The aim of this paper is to prove the global property of
120: embeddedness for
121: the square grid circle patterns corresponding to $z^{\gamma}$ (which
122: was conjectured in \cite{AB}).
123:
124: To visualize the analogy between Schramm's circle patterns and
125: conformal maps, consider regular patterns composed of unit circles
126: and suppose that the radii are being deformed so as to preserve
127: the orthogonality of neighboring circles and the tangency of
128: half-neighboring ones. Discrete maps taking intersection points of
129: the unit circles of the standard regular patterns to the
130: respective points of the deformed patterns mimic classical
131: holomorphic functions, the deformed radii being analogous to
132: $|f'(z)|$ (see Fig. \ref{CPMap}).
133:
134: \begin{figure}[th]
135: \begin{center}
136: \epsfig{file=PreImage.eps,width=45mm}
137: \begin{picture}(20,50)
138: \put(2,22){\vector(1,0){15}} \put(7,27){ \it \huge f}
139: \end{picture}
140: \epsfig{file=Image.eps,width=50mm} \caption{Schramm's circle
141: patterns as discrete conformal map. Shown is the discrete version
142: of the holomorphic mapping $z^{3/2}$.} \label{CPMap}
143: \end{center}
144: \end{figure}
145:
146: It is easy to show that the lattice comprised of the centers of
147: circles of Schramm's pattern and their intersection
148:
149: points is a special discrete conformal mapping (see Definition
150: \ref{DCM} below). The latter were introduced in \cite{BPdis} in
151: the frames of discrete integrable geometry, originally without any
152: relation to circle patterns.
153:
154: \begin{definition} A map $f\ : \ {\bf Z^2\ \rightarrow
155: \ {\bf R^2}={\bf C}}$ is called a discrete conformal map if all
156: its elementary quadrilaterals are conformal squares, i.e., their
157: cross-ratios are equal to -1: $$
158: q(f_{n,m},f_{n+1,m},f_{n+1,m+1},f_{n,m+1}):= $$
159: \begin{equation}
160: \frac{(f_{n,m}-f_{n+1,m})(f_{n+1,m+1}-f_{n,m+1})}
161: {(f_{n+1,m}-f_{n+1,m+1})(f_{n,m+1}-f_{n,m})}=-1 \label{q}
162: \end{equation}
163: \label{DCM}
164: \end{definition}
165: This definition is motivated by the following properties:\\ 1) it
166: is M\"obius invariant,\\ 2) a smooth map $f:\ D\ \subset {\bf C}
167: \to {\bf C}$ is conformal (holomorphic or antiholomorphic) if and
168: only if $$ \lim _{\epsilon \to 0}q(f(x,y),f(x+\epsilon
169: ,y)f(x+\epsilon ,y+\epsilon )f(x,y+\epsilon ))=-1 $$ for all
170: $(x,y)\in D$. For some examples see \cite{BPdis},\cite{TH}. A
171: naive method to construct a discrete analogue of the function
172: $f(z)=z^{\gamma }$ is to start with $f_{n,0}=n^{\gamma }, \ n\ge
173: 0$, $f_{0,m}=(im)^{\gamma }, \ m\ge 0$ and then compute $f_{n,m}$
174: for any $n,m > 0$ using equation (\ref{q}). But so determined map
175: has a behavior which is far from that of usual holomorphic maps.
176: Different elementary quadrilaterals overlap (see the left lattice
177: in Fig. \ref{correct-wrong}).
178: \begin{figure}[th]
179: \begin{center}
180: \epsfig{file=zalpha_wrong.eps,width=50mm}
181: \epsfig{file=zalpha.eps,width=50mm} \caption{Two discrete
182: conformal maps with close initial data $n=0, m=0$. The second
183: lattice describes a discrete version of the holomorphic mapping
184: $z^{2/3}$.} \label{correct-wrong}
185: \end{center}
186: \end{figure}
187: \begin{definition}
188: A discrete conformal map $f_{n,m}$ is called embedded if inner
189: parts of different elementary quadrilaterals
190: $(f_{n,m},f_{n+1,m},f_{n+1,m+1},f_{n,m+1})$ do not intersect.
191: \end{definition}
192: \noindent The condition for discrete conformal map to be embedded
193: can be relaxed as follows.
194: \begin{definition}
195: A discrete conformal map $f_{n,m}$ is called an immersion if inner
196: parts of adjacent elementary quadrilaterals
197: $(f_{n,m},f_{n+1,m},f_{n+1,m+1},f_{n,m+1})$ are disjoint.
198: \end{definition}
199: To construct an embedded discrete analogue of $z^{\gamma },$ which
200: is the right lattice presented in Fig. \ref{correct-wrong}, a more
201: complicated approach is needed. Equation (\ref{q}) can be
202: supplemented with the following nonautonomous constraint:
203: \begin{equation}
204: \gamma f_{n,m}=2n\frac{(f_{n+1,m}-f_{n,m})(f_{n,m}-f_{n-1,m})}
205: {(f_{n+1,m}-f_{n-1,m})}+2m\frac{(f_{n,m+1}-f_{n,m})(f_{n,m}-f_{n,m-1})}
206: {(f_{n,m+1}-f_{n,m-1})}, \label{c}
207: \end{equation}
208: which plays a crucial role in this paper. This constraint, as well
209: as its compatibility with (\ref{q}), is derived from some
210: monodromy problem (see \cite{AB}). Let us assume $0<\gamma < 2$
211: and denote ${\bf Z^2_{+}}=\{ (n,m) \in {\bf Z^2}: n,m \ge 0 \}.$
212: Motivated by the asymptotics of the constraint (\ref{c}) at $n,m
213: \rightarrow \infty$ and the properties $$ z^\gamma({\bf R_+}) \in
214: {\bf R_+}, \ \ z^{\gamma}(i{\bf R_+}) \in e^{\gamma \pi i/2}{\bf
215: R_+} $$ of the holomorphic mapping $z^{\gamma}$ we use the
216: following definition \cite{BPD} of the "discrete" $z^{\gamma}$.
217: \begin{definition}
218: The discrete conformal map $Z^{\gamma}\ : \ {\bf Z^2_+\
219: \rightarrow \ {\bf C}}, \ 0<\gamma < 2\ $ is the solution of
220: (\ref{q}),(\ref{c}) with the initial conditions
221: \begin{equation}
222: Z^{\gamma}(0,0)=0, \ \ Z^{\gamma}(1,0)=1, \ \
223: Z^{\gamma}(0,1)=e^{\gamma \pi i/2}. \label{initial}
224: \end{equation}
225: \label{def}
226: \end{definition}
227: Obviously, $Z^{\gamma}(n,0)\in {\bf R_{+}}$ and
228: $Z^{\gamma}(0,m)\in e^{\gamma \pi i/2} ({\bf R_{+}})$ for any $
229: n,m \in {\bf N}$ .\\
230: Fig. \ref{correct-wrong} suggests that $Z^{\gamma }$ is
231: embedded. The corresponding theorem is the main result of this
232: paper.
233: \begin{theorem}
234: The discrete map $Z^{\gamma }$ for $0<\gamma <2$ is embedded.
235: \label{main}
236: \end{theorem}
237: The proof is based on the analysis of geometric and algebraic
238: properties of the corresponding lattices. In Section 2 a brief
239: review of the necessary results from \cite{AB} is given. It is
240: shown that $Z^{\gamma }$ corresponds to a circle pattern of
241: Schramm's type. (The circle pattern corresponding to $Z^{2/3}$ is
242: presented in Fig. \ref{f.2over3} )
243: \begin{figure}[t]
244: \begin{center}
245: \epsfig{file=2over3.eps, width=50mm}
246: \end{center}
247: \caption{Schramm's circle pattern corresponding to $Z^{2/ 3}$}
248: \label{f.2over3}
249: \end{figure}
250: Next, analyzing the equations for radii of the circles we show
251: that in order to prove that $Z^{\gamma }$ is embedded it is enough
252: to establish a special property of the separatrix solutions
253: $P_{N,M},~Q_{N,M}$
254: of
255: the following infinite sequence of systems of ordinary difference
256: equations of Painlev\'e type ($N$ labels the system): $$
257: Q_{N,M+1}=\frac{
258: (M+N)P_{N,M}(P_{N,M}-Q_{N,M}^2)-(M-N)Q_{N,M}^2(1+P_{N,M})
259: }{Q_{N,M}( (M+N)(Q_{N,M}^2-P_{N,M})-(M-N)P_{N,M}(1+P_{N,M}) )} $$
260: $$ P_{N,M+1}=\frac{ (2M+\gamma )P_{N,M} + (2N+\gamma
261: )Q_{N,M}Q_{N,M+1} }{ (2(N+1)-\gamma )P_{N,M} + (2(M+1)-\gamma
262: )Q_{N,M}Q_{N,M+1} } $$ Namely, it is shown that $Z^{\gamma }$ is
263: embedded if the infinite sequence of solutions $Q_{N,M},P_{N,M}$
264: with special initial data $Q_{N,N},P_{N,N}$
265: is subject to $$ (\gamma-1)(Q_{N,M}^2 - P_{N,M})\ge 0, \
266: Q_{N,M}>0, \ P_{N,M}>0 $$ The existence and uniqueness of the
267: corresponding initial data are given in Section 3.
268:
269: For $N=0$ the system for $Q_{N,M},P_{N,M}$ reduces to the special
270: case of discrete Painlev\'e equation dPII
271: \begin{equation}
272: \label{dPII}
273: (n+1)(x_n^2-1)\left(\frac{x_{n+1}-ix_n}{i+x_nx_{n+1}}\right)-
274: n(x_n^2+1)\left(\frac{x_{n-1}+ix_n}{i+x_{n-1}x_n}\right) =
275: \gamma x_n
276: \end{equation}
277: (see \cite{FIK}, \cite{NRGO} for more examples). In \cite{AB} this
278: equation was the main tool to prove the immersion of $Z^{\gamma}$:
279: it was shown that $Z^{\gamma }$ is immersed if
280: the unitary solution $x_n=e^{i\alpha_n}$ of this equation with
281: $x_0=e^{i\gamma \pi/4}$ lies in the sector $0<\alpha _n < \pi/2.$
282:
283: Similar problems have been studied in the frames of the
284: isomonodromic deformation method \cite{IN}. In particular,
285: connection formulas were derived. These formulas describe the
286: asymptotics of solutions for $n \to \infty$ as a function of
287: initial conditions (see in particular \cite{FIK}). These methods
288: seem to be insufficient for our purposes since we need to control
289: the behavior of solutions for finite $n$'s as well. The geometric
290: origin of our equations permits us to prove the abovementioned
291: properties by purely geometric methods. To illustrate the
292: difference between the immersed and embedded discrete conformal
293: maps, let us imagine that the elementary quadrilaterals of the map
294: are made of elastic inextensible material and glued along the
295: corresponding edges to produce
296: a
297: surface with a border. If this surface is immersed it is locally
298: flat. Being dropped down it will not have folds. At first sight it
299: seems to be sufficient to give embeddedness, provided $Z_{n,0}\to
300: \infty$ and $Z_{0,m}\to \infty$ as $n \to \infty$ (which follows
301: from
302: $
303: Z^{\gamma }_{n,0}=\frac{2c(\gamma )}{\gamma}\left(\frac {n}{2}
304: \right)^{\gamma }\left(1+O\left(\frac{1}{n^2} \right) \right), \ \
305: n \to \infty,
306: $
307: see \cite{AB}). But a surface with such properties still may have
308: some limit curve with self-intersections thus giving overlapping
309: quadrilaterals. Hypothetical example of such a surface is shown in
310: Fig. \ref{surface}.
311: \begin{figure}[ht]
312: \begin{center}
313: \epsfig{file=surface.eps, width=50mm}
314: \end{center}
315: \caption{Surface glued of quadrilaterals of immersed but
316: non-embedded discrete map. } \label{surface}
317: \end{figure}
318: \section{Circle
319: patterns and $Z^{\gamma}$} Let us recall the necessary results
320: from \cite{AB}. $Z^{\gamma}$ of Definition \ref{def} determines a
321: special case of circle patters with the combinatorics of the
322: square grid as defined by Schramm in \cite{Schramm}. Indeed,
323: discrete $f_{n,m}$ satisfying (\ref{q}) and (\ref{c}) with initial
324: data $f_{0,0}=0$, $f_{1,0}=1$, $f_{0,1}=e^{i \alpha }$ has the
325: equidistant property $$ f_{2n,0}-f_{2n-1,0}=f_{2n+1,0}-f_{2n,0},\
326: \ f_{0,2m}-f_{0,2m-1}=f_{0,2m+1}-f_{0,2m} $$ for any $n\ge1$,
327: $m\ge1$. \label{equi} \noindent Given initial $f_{0,0}$, $f_{0,1}$
328: and $f_{1,0}$ the constraint (\ref{c}) allows one to compute
329: $f_{n,0}$ and $f_{0,m}$ for all $n,m \ge 1.$ Now using equation
330: (\ref{q}) one can successively compute $f_{n,m}$ for any $n,m \in
331: {\bf N}$. Observe that if
332: $|f_{n+1,m}-f_{n,m}|=|f_{n,m+1}-f_{n,m}|$
333: then the quadrilateral
334: $(f_{n,m},f_{n+1,m},f_{n+1,m+1},f_{n,m+1})$ is of the kite form --
335: it is inscribed in a circle and is symmetric with respect to the
336: diameter of the circle $[f_{n,m}, f_{n+1,m+1}].$ If the angle at
337: the vertex $f_{n,m}$ is $\pi /2$ then the quadrilateral
338: $(f_{n,m},f_{n+1,m},f_{n+1,m+1},f_{n,m+1})$ is of the kite form
339: too. In this case the quadrilateral is symmetric with respect to
340: its diagonal $[f_{n,m+1}, f_{n+1,m}]$.
341:
342: \begin{proposition}\cite{AB}
343: Let $f_{n,m}$ satisfy (\ref{q}) and (\ref{c}) in ${\bf Z^2_+}$
344: with initial data $f_{0,0}=0$, $f_{1,0}=1$, $f_{0,1}=e^{i \alpha
345: }.$ Then all the elementary quadrilaterals $(f_{n,m},
346: f_{n+1,m},f_{n+1,m+1},f_{n,m+1})$ are of the kite form. All edges
347: at the vertex $f_{n,m}$ with $n+m=0 \ ({\rm mod} \ 2)$ are of the
348: same length $$
349: |f_{n+1,m}-f_{n,m}|=|f_{n,m+1}-f_{n,m}|=|f_{n-1,m}-f_{n,m}|=|f_{n,m-1}-f_{n,m}|.
350: $$ All angles between the neighboring edges at the vertex
351: $f_{n,m}$ with $n+m=1 \ ({\rm mod} \ 2)$ are equal to $\pi /2.$
352: \label{kite}
353: \end{proposition}
354: \noindent Proposition \ref{kite} implies that for any $n,m: \
355: n+m=0 \ ({\rm mod }\ 2)$ the points $f_{n+1,m},$ $f_{n,m+1},$
356: $f_{n-1,m},$ $f_{n,m-1}$ lie on a circle with the center
357: $f_{n,m}$.
358: \begin{corollary}\cite{AB}
359: The circumscribed circles of the quadrilaterals $(f_{n-1,m},
360: f_{n,m-1},f_{n+1,m},f_{n,m+1})$ with $n+m=0 \ ({\rm mod} \ 2)$
361: form a circle pattern of Schramm type (see \cite{Schramm}), i.e.
362: the circles of neighboring quadrilaterals intersect orthogonally
363: and the circles of
364: half-neighboring
365: quadrilaterals with common vertex are tangent (see Fig. 3).
366: \end{corollary}
367:
368: Consider the sublattice $\{n,m: \ n+m=0 \ ({\rm mod} \ 2)\}$ and
369: denote by $\bf V$ its quadrant $$ {\bf V}=\{z=N+iM:\ N,M \in {\bf
370: Z^2}, M \ge |N| \}, $$ where $$ N=(n-m)/2, \ \ M=(n+m)/2. $$
371: We
372: will use complex labels $z=N+iM$ for this sublattice. Denote by
373: $C(z)$ the circle of the radius
374:
375: \begin{equation}
376: R(z)=|f_{n,m}-f_{n+1,m}|=|f_{n,m}-f_{n,m+1}|=|f_{n,m}-f_{n-1,m}|=|f_{n,m}-f_{n,m-1}|
377: \label{rmap}
378: \end{equation}
379: with the center at $f_{N+M,M-N}=f_{n,m}.$ From Proposition
380: \ref{kite} it follows that any two circles $C(z)$, $C(z')$ with
381: $|z-z'|=1$ intersect orthogonally and any two circles $C(z)$,
382: $C(z')$ with $|z-z'|=\sqrt{2}$ are tangent.
383: \\
384: Let $\{C(z)\}, \ z\in {\bf V}$ be a circle pattern of Schramm
385: type on the complex plane. Define $f_{n,m}: {\bf Z^2_+ \to C}$ as
386: follows:\\ a) if $n+m=0 \ ({\rm mod} \ 2)$ then $f_{n,m}$ is the
387: center of $C(\frac{n-m}{2}+i\frac{n+m}{2}),$\\ b) if $n+m=1 \
388: ({\rm mod} \ 2)$ then
389: $
390: f_{n,m}:=C(\frac{n-m-1}{2}+i\frac{n+m-1}{2}) \cap
391: C(\frac{n-m+1}{2}+i\frac{n+m+1}{2})=
392: C(\frac{n-m+1}{2}+i\frac{n+m-1}{2}) \cap
393: C(\frac{n-m-1}{2}+i\frac{n+m+1}{2}).
394: $
395: Since all elementary quadrilaterals $(f_{n,m},
396: f_{n+1,m},f_{n+1,m+1},f_{n,m+1})$ are of the kite form equation
397: (\ref{q}) is satisfied automatically. In what follows the function
398: $f_{n,m},$ defined as above by a) and b) is called {\it a discrete
399: conformal map corresponding to the circle pattern $\{C(z)\}$ .}
400: \begin{theorem}\cite{AB}
401: Let $f_{n,m}$ satisfying (\ref{q}) and (\ref{c}) with initial data
402: $f_{0,0}=0$, $f_{1,0}=1$, $f_{0,1}=e^{i \alpha }$ be an immersion,
403: then $R(z)$ defined by (\ref{rmap}) satisfies the following
404: equations:
405: \begin{equation}
406: \begin{array}{l}
407: R(z)R(z+1)(-2M-\gamma )+R(z+1)R(z+1+i)(2(N+1)-\gamma )+\\ \qquad
408: R(z+1+i)R(z+i)(2(M+1)-\gamma )+R(z+i)R(z)(-2N-\gamma )=0,
409: \end{array}
410: \label{square}
411: \end{equation}
412: for $z\in {\bf V}_l:={\bf V}\cup \{-N+i(N-1)|N\in {\bf N}\}$ and
413: \begin{equation}
414: \begin{array}{l}
415: (N+M)(R(z)^2-R(z+1)R(z-i))(R(z+i)+R(z+1))+\\ \qquad
416: (M-N)(R(z)^2-R(z+i)R(z+1))(R(z+1)+R(z-i))=0,
417: \end{array}
418: \label{Ri}
419: \end{equation}
420: for $z\in {\bf V}_{rint}:={\bf V}\backslash \{\pm N+iN|N\in {\bf
421: N}\}.$\\ Conversely let $R(z): {\bf V} \to {\bf R_+}$ satisfy
422: (\ref{square}) for $z\in {\bf V}_l$ and (\ref{Ri}) for $z\in {\bf
423: V}_{rint}.$ Then $R(z)$ define an immersed circle packing with the
424: combinatorics of the square grid. The corresponding discrete
425: conformal map $f_{n,m}$ is an immersion and satisfies (\ref{c}).
426: \label{eqforR}
427: \end{theorem}
428: \noindent From the initial condition (\ref{initial}) we have
429: \begin{equation}
430: R(0)=1, \ \ \ R(i)=\tan \frac{\gamma \pi}{4}. \label{Rinitial}
431: \end{equation}
432: Equation (\ref{square}) at $z=N+iN$ and $z=-N+i(N-1),\ N\in {\bf
433: N}$ reads as
434: \begin{equation}
435: R(\pm (N+1)+i(N+1))=\frac{2N+\gamma}{2(N+1)-\gamma }R(\pm N+iN)
436: \label{recR}
437: \end{equation}
438: and defines $R(\pm N+iN)$ for all $N \in {\bf N}.$ Now equations
439: (\ref{square},\ref{Ri}) determine $R(z)$ for any other $z\in {\bf
440: V}.$ Besides $R(z)$ satisfy
441: (\ref{Ri}) at
442: $z=-N+iN, \ N \in {\bf N}$ which reads
443: $$
444: R(-N+1+iN)R(-N+i(N+1))=R^2(-N+iN).
445: \label{line}
446: $$
447: By symmetry one gets $$ R(N-1+iN)R(N+i(N+1))=R^2(N+iN). $$
448: This equation
449: allows to compute $R(N+i(N+1))$. (Moreover, it implies that the
450: center $O$ of $C(N+iN)$ and the points $A=C(N+iN) \cap C(N-1+iN)
451: \backslash C(N-1+iN) \cap C(N+i(N+1))$
452: and
453: $B=C(N+iN) \cap C(N+i(N+1)) \backslash C(N-1+iN) \cap C(N+i(N+1))$
454: are collinear so the points $f_{n,0}$ lie on a straight line.) To
455: prove Theorem \ref{main} we consider more general initial
456: condition:
457: \begin{equation}
458: R(0)=1, \ \ \ R(i)=a\tan \frac{\gamma \pi}{4}. \label{Rainitial}
459: \end{equation}
460: Now the radii $R(N+iN),R(N+i(N+1)$ can be represented in terms of
461: $\Gamma$-~function:
462: \begin{equation}
463: R(N+iN)=c(\gamma )\frac{\Gamma(N+\gamma/2)}{\Gamma(N+1-\gamma
464: /2)}, \ \
465: \mbox{where }
466: c(\gamma )=\frac{\gamma \Gamma(1-\gamma /2)}{2\Gamma(1+\gamma
467: /2)}, \label{Rline}
468: \end{equation}
469: \begin{equation}
470: R(N+i(N+1))=\left( a\tan \frac {\gamma \pi}{4} \right)^{(-1)^N}
471: \left(\frac{ (2(N-1)+\gamma )(2(N-3)+\gamma )(2(N-5)+\gamma )...}
472: {(2N-\gamma )(2(N-2)-\gamma )(2(N-4)-\gamma )... } \right)^2.
473: \label{Ronline}
474: \end{equation}
475: Theorem \ref{eqforR} allows us to reformulate the property of the
476: circle lattice to be immersed completely in terms of the system
477: (\ref{square},\ref{Ri}). Namely to prove that $Z^{\gamma}$ is an
478: immersion one should show that the solution of the system
479: (\ref{square},\ref{Ri}) with initial data (\ref{Rinitial}) is
480: positive for all $z \in {\bf V}$ (see \cite{AB}). To prove Theorem
481: \ref{main} one need more subtle property of this solution.
482: \begin{theorem}
483: If for a solution R(z) of (\ref{square}, \ref{Ri}) with $\gamma
484: \ne 1$ and initial conditions (\ref{Rinitial}) holds
485: \begin{equation}
486: R(z)>0, \
487: \ (\gamma -1)(R(z)^2 - R(z-i)R(z+1))\ge 0 \label{sign}
488: \end{equation}
489: in ${\bf V}_{int}$, then the corresponding discrete conformal map
490: is embedded. \label{convex}
491: \end{theorem}
492: \noindent {\it Proof:} Since $R(z)>0$ the corresponding discrete
493: conformal map is immersion (see \cite{Schramm,AB}).
494: Consider
495: piecewise linear curve $\Gamma _n$ formed by segments
496: $[f_{n,m},f_{n,m+1}]$ where $n>0$ and $0\le m \le n-1$ and the
497: vector ${\bf v}_n(m)={\bf (f_{n,m}f_{n,m+1})}$ along this curve.
498: Due to Proposition \ref{kite} this vector
499: rotates only in vertices with
500: $n+m=0 \ ({\rm mod} \ 2)$ as $m$ increases along the curve. The
501: sign of the rotation angle $\theta _n(m)$, where $-\pi < \theta
502: _n(m) < \pi$, $0<m<n$ is defined by the sign of expression
503: $R(z)^2-R(z+i)R(z+1)$ (note that there is no rotation if this
504: expression vanishes), where
505: $z=(n-m)/2+i(n+m)/2$ is a label
506: for the circle with the center in $f_{n,m}$. If $n+m=1\ ({\rm mod
507: \ 2})$ define $\theta _n(m)=0$. Now the theorem hypothesis and
508: equation (\ref{Ri}) imply that
509: the vector
510: ${\bf v}_n(m)$ rotates with increasing $m$ in the same direction
511: for all $n$, and namely, clockwise for $\gamma <1$ and
512: counterclockwise for $\gamma >1$. Consider the sector $B:=\{
513: z=re^{i\varphi}:r\ge 0,\ 0\le \varphi \le \gamma \pi /4 \}$. The
514: terminal
515: points of the curves $\Gamma _n$ lie
516: on the sector border.
517: \begin{lemma}
518: For the curve $\Gamma _n$ holds:
519: \begin{equation}
520: \left |\sum_{m=1}^{n-1} \theta _n(m) \right | < \frac{
521: \pi}{4}(1+|1-\gamma |) \label{ineq}
522: \end{equation}
523: \label{rot}
524: \end{lemma}
525: \noindent {\it Proof of Lemma \ref{rot}}: Let us prove the
526: inequality (\ref{ineq}) for $1 < \gamma <2$ by induction for $n$.
527: For $n=1$ the inequality is obviously true since the curve $\Gamma
528: _1$ is a segment perpendicular to $\bf R_+$.
529: Define the angle $\alpha _n(m)$ between
530: $i{\bf R_+}$ and the vector ${\bf v}_n(m)$
531: by
532: $f_{n,m+1}-f_{n,m}=e^{i(\alpha _n(m)+\pi/2)}|f_{n,m+1}-f_{n,m}|$,
533: where $0 \le \alpha _n(m) <2\pi$, $0\le m <n$. Then
534: $\sum_{m=1}^{l} \theta _n(m)=\alpha _n(l)-\alpha _n(0)+2\pi
535: k_n(l)$ for some positive integer $k_n(l)$ increasing with $l$.
536: Note that $\alpha _n(0) < \pi/2 $, which easily follows from
537: (\ref{Rline},\ref{Ronline})), and $\alpha _n(n-1) < (\frac{\gamma
538: \pi}{4}+\frac{\pi}{2})-\frac{\pi}{2}= \frac{\gamma \pi}{4} $ since
539: for immersed $Z^{\gamma }$ the angle between
540: the vector ${\bf v}_n(n-1)$ and
541: $ e^{i\gamma \pi/4}{\bf R_+}$ is less then $\frac{\pi}{2}$. Let
542: (\ref{ineq}) holds for $n>1$: $\left |\sum_{m=1}^{n-1} \theta
543: _n(m) \right |=\sum_{m=1}^{n-1} \theta _n(m) < \frac{\gamma
544: \pi}{4}$ (all $\theta _n(m)$ are positive for $1 < \gamma <2$).
545: That implies $k_n(l)$=0, since $k_n(l)=(\sum_{m=1}^{l} \theta
546: _n(m)-\alpha _n(l)+\alpha _n(0))/2\pi \le (\sum_{m=1}^{l} \theta
547: _n(m)+|\alpha _n(l)|+|\alpha _n(0)|)/2\pi < (\gamma \pi /4
548: +\gamma \pi /4+\pi /2)/2\pi < 1$ and $k_n(l)$ is integer. Let
549: $\alpha _{n+1}(l)=\alpha _n(l)+\sigma _n(l)$. All elementary
550: quadrilaterals are of the kite form therefore $|\sigma
551: _n(l)|<\pi/2$. Let us prove, that $k_{n+1}(m)=0$ for $0\le m \le
552: n+1$. Obviously $k_{n+1}(0)=0$. Assume $k_{n+1}(l)=0$ but
553: $k_{n+1}(l+1)>0$. The increment of l.h.s. of $$ \sum_{m=1}^{l}
554: \theta _{n+1}(m)=\alpha _{n+1}(l)-\alpha _{n+1}(0)+2\pi k_{n+1}(l)
555: $$ as $l\to l+1$ is $\theta _{n+1}(l+1)<\pi$. The increment of
556: r.h.s. is no less than $2\pi +\alpha _{n+1}(l+1)-\alpha _{n+1}(l)
557: \ge 2\pi-\alpha _{n+1}(l) \ge 2\pi -\alpha _n(l)-|\sigma _n(l)|
558: > 2\pi -\gamma \pi /4 -\pi/2 > \pi.$ The
559: obtained contradiction gives $k_{n+1}(l)=0$ and $\sum_{m=1}^{n}
560: \theta _{n+1}(m)=\alpha _{n+1}(n)-\alpha _{n+1}(0)
561: \le \alpha _{n+1}(n) <
562: \frac{\gamma \pi}{4}.$ Lemma \ref{rot} is proved.
563:
564: \smallskip
565: The obvious corollary of Lemma \ref{rot} is that the curve $\Gamma
566: _n$ has no self-intersection and lies in the sector $B$
567: since the rotation of
568: the vector ${\bf v}_n(m)$ along the curve is less then $\gamma
569: \pi/4<\pi/2$. Each such curve cuts the sector $B$ into a finite
570: part and an infinite part. Since the curve $\Gamma _n$ is convex
571: and the borders of all elementary quadrilaterals
572: $(f_{n,m},f_{n+1,m},f_{n+1,m+1},f_{n,m+1})$ for embedded
573: $Z^{\gamma }$ have the positive orientation
574: the
575: segments of the curve $\Gamma _{n+1}$ lie in the infinite part.
576: Now the induction in $n$ completes the proof of Theorem
577: \ref{convex} for $1<\gamma <2$. The proof for
578: $0<\gamma <1$ is similar. The differences
579: are that $\theta _n(m)$ is not positive, the angle $\alpha $ is
580: naturally defined as negative: $-2\pi <\alpha _n(m) <0$, so that
581: $-\pi/2
582: <\alpha _n(0) \le 0 $ and $\frac{\gamma \pi}{4}(2-\gamma) <\alpha
583: _n(n-1) < 0$. Details are left to the reader.
584: \begin{corollary}
585: $$ \lim _{n\to \infty}Z^{\gamma} _{n,m}=\infty, \ \ \lim _{m\to
586: \infty}Z^{\gamma} _{n,m}=\infty. $$
587: \end{corollary}
588: Since the terminal
589: points of the
590: curves $\Gamma _n$ lie on the sector border the proof easily
591: follows from convexity of the curves $\Gamma _n$, inequality
592: (\ref{ineq}) and from $$ \lim _{n\to \infty}Z^{\gamma}
593: _{n,0}=\infty. $$
594:
595: \section{
596: $Z^{\gamma}$ and discrete Painlev\'e equations} Let $R(z)$ be a
597: solution of (\ref{square},\ref{Ri}) with initial condition
598: (\ref{Rinitial}). For $z\in {\bf V}_{int}$ define
599: $P_{N,M}=P(z)=\frac{R(z+1)}{R(z-i)}$,
600: $Q_{N,M}=Q(z)=\frac{R(z)}{R(z-i)}$. Then equations
601: (\ref{square},\ref{Ri}) are rewritten as follows
602: \begin{equation}
603: Q_{N,M+1}=\frac{
604: (M+N)P_{N,M}(P_{N,M}-Q_{N,M}^2)-(M-N)Q_{N,M}^2(1+P_{N,M})
605: }{Q_{N,M}( (M+N)(Q_{N,M}^2-P_{N,M})-(M-N)P_{N,M}(1+P_{N,M}) )}
606: \label{QPainleve}
607: \end{equation}
608: \begin{equation}
609: P_{N,M+1}=\frac{ (2M+\gamma )P_{N,M} + (2N+\gamma
610: )Q_{N,M}Q_{N,M+1} }{ (2(N+1)-\gamma )P_{N,M} + (2(M+1)-\gamma
611: )Q_{N,M}Q_{N,M+1} } \label{PPainleve}
612: \end{equation}
613: The property (\ref{sign}) for (\ref{QPainleve},\ref{PPainleve})
614: reads as
615: \begin{equation}
616: (\gamma-1)(Q_{N,M}^2 - P_{N,M})\ge 0, \ Q_{N,M}>0, \ P_{N,M}>0.
617: \label{PQsign}
618: \end{equation}
619: Equations (\ref{QPainleve},\ref{PPainleve}) can be considered as a
620: dynamical system for variable $M$,
621: the expressions (\ref{Rline},\ref{Ronline}) defining
622: initial conditions $P_{N,N+1},Q_{N,N+1}$ for
623: (\ref{QPainleve},\ref{PPainleve})
624: as
625: functions of $a,N$, only $Q_{N,N+1}$ being dependent on $a$.
626: \begin{theorem}
627: There exists such $a>0$ that for the solutions $R(z)$ of
628: (\ref{square},\ref{Ri}) with initial conditions (\ref{Rainitial})
629: holds (\ref{sign}). \label{existence}
630: \end{theorem}
631: \noindent {\it Proof:} Due to the following Lemma it is sufficient
632: to prove (\ref{sign}) only for $0<\gamma<1$.
633: \begin{lemma}
634: If $R(z)$ is a solution of (\ref{square},\ref{Ri}) for $\gamma$
635: then $1/R(z)$ is a solution of (\ref{square},\ref{Ri}) for $\tilde
636: \gamma =2-\gamma$.
637: \end{lemma}
638: Lemma is proved by straightforward computation.
639:
640: \smallskip
641: Let $0<\gamma <1$ and $(P_{N,M},Q_{N,M})$ correspond to the
642: solution of (\ref{square},\ref{Ri}) defined by initial conditions
643: (\ref{Rainitial}). Define the real function $F(P)$ for $P\in {\bf
644: R_+}$:
645: $$F(P)=\sqrt{P} \ {\rm for}
646: \
647: 0\le P \le 1, \ F(P)=1\ {\rm for} \ 1\le P. $$ Designate $$
648: D_{u}:=\{(P,Q): P > 0, Q > F(P) \},\ \ D_{d}:=\{(P,Q): Q < 0 \},
649: $$ $$
650: D_{0}:=\{(P,Q): P > 0, 0 \le
651: Q \le F(P) \}, \ \ D_f:=\{ (P,Q): P\le 0, Q\ge 0 \}. $$ Now
652: define the infinite sequences $\{q_n\},\{p_n \},\ n\in {\bf N}$ as
653: follows: $$ \{q_n(a)\}:=\{ Q_{0,1},Q_{0,2},
654: Q_{1,2},Q_{0,3},Q_{1,3},Q_{2,3},...,Q_{0,M},Q_{1,M},...,Q_{M-1,M},...
655: \}, $$ $$ \{p_n(a)\}:=\{ P_{0,1},P_{0,2},
656: P_{1,2},P_{0,3},P_{1,3},P_{2,3},...,P_{0,M},P_{1,M},...,P_{M-1,M},...
657: \}. $$ and the sets $$ A_u(n):=\{a\in {\bf R_+}:
658: (p_n(a),q_n(a))\in D_u, (p_k(a),q_k(a))\in D_0 \ \forall \ 0< k <n
659: \}, $$ $$ A_d(n):=\{a\in {\bf R_+}: (p_n(a),q_n(a))\in D_d,
660: (p_k(a),q_k(a))\in D_0 \ \forall \ 0< k <n \}. $$ $A_u(n)$ and
661: $A_d(n)$ are open sets since the denominators of
662: (\ref{QPainleve},\ref{PPainleve}) do not vanish in $D_0$. Indeed,
663: the curve $Q^2=P+\frac{(M-N)}{(M+N)}P(P+1)$ lies outside $D_0.$
664: Moreover, direct computation shows that
665: $A_u(1)\ne \emptyset$ and $A_d(2)\ne
666: \emptyset,$ therefore the sets $$ A_u:=\cup A_u(k), \ \ \
667: A_d:=\cup A_d(k) $$ are not empty. Finally, define $$ A_0:=\{a\in
668: {\bf R_+}: (p_n(a),q_n(a))\in D_0, \ \forall \ n \in{\bf N} \}.
669: $$ Note that $A_0,A_u,A_d$ are mutually disjoint and the sequences
670: $\{p_n\},\{q_n\}$ is so constructed that
671: \begin{equation}
672: \label{union} {\bf R_+}=A_0\cup A_u \cup A_d.
673: \end{equation}
674: Indeed $(P_{N,M},Q_{N,M})$ can not jump from $D_0$ into $D_f$ in
675: one step $M \to M+1$ since $P_{N,M+1}$ is positive for positive
676: $P_{N,M},Q_{N,M},Q_{N,M+1}$.
677: The relation (\ref{union}) would be impossible for
678: $A_0=\emptyset$, since the connected set ${\bf R_+}$ can not be
679: covered by two open
680: disjoint nonempty
681: subsets $A_u$ and $A_d$, therefore $A_0\ne \emptyset.$ Q.E.D.
682:
683: \begin{proposition}
684: \label{true_a} The set $A_0$ consists of only one element, namely,
685: $A_0=\{1\}$.
686: \end{proposition}
687:
688: \noindent {\it Proof:} A positive solution $R(z)$ of
689: (\ref{square},\ref{Ri})
690: provides an immersed discrete conformal map corresponding
691: to the circle patterns
692: with radii $R(z)$. It was proven in \cite{AB}
693: that the initial value (\ref{Rainitial}) for immersed $Z^{\gamma}$
694: consists of only one element $a=1$. Q.E.D.
695:
696: \medskip
697:
698: \noindent {\it Proof of Theorem \ref{main}:} combining Theorems
699: \ref{convex} and \ref{existence} with Proposition \ref{true_a} one
700: easily deduces Theorem \ref{main}.
701:
702: \section{Concluding remarks}
703:
704: The approach suggested in this paper can be applied to prove the
705: embeddedness of the circle patterns corresponding to discrete
706: $Z^2$ and ${\rm Log}$. (These patterns were proved to be immersed
707: \cite{AB}.) Slightly modified, it seems to
708: be also applicable to show global properties of the discrete $Z^K$
709: proposed in \cite{AB} for natural $K>2$. Details will be discussed
710: elsewhere.
711:
712: \section{Acknowledgements}
713:
714: The author would like to thank A.Bobenko and Yu.Suris for useful
715: discussions. This research was supported by the EPSRC grant No
716: Gr/N30941.
717:
718: \begin{thebibliography}{99}
719:
720: \addcontentsline{toc}{section}{References}
721:
722: \bibitem{AB}
723: S.I.Agafonov, A.I.Bobenko, {\it Discrete $Z^\gamma$ and Painlev\'e
724: equations,} Internat. Math. Res.
725: Notices, {\bf 4} (2000), 165-193
726:
727: \bibitem{BS}
728:
729: A.F.Beardon, K.Stephenson {\it The uniformization theorem for
730: circle packings,} Indiana Univ. Math. J., {\bf 39:4} (1990),
731: 1383-1425
732:
733:
734: \bibitem{BHConf} A.I.Bobenko, T.Hoffman,
735: {\it Conformally symmetric circle packings. A generalization of
736: Doyle spirals,} Experimental Math., {\bf 10:1} (2001), 141-150
737:
738: \bibitem{BHS} A.I.Bobenko, T.Hoffman,
739: Yu.B.Suris, {\it Hexagonal circle patterns and integrable systems:
740:
741: Patterns with the multi-ratio property and Lax equations on the
742: regular triangular lattice,} Internat. Math. Res.
743: Notices, {\bf 3} (2002),
744: 111-164
745:
746: \bibitem{BH} A.I.Bobenko, T.Hoffman,
747: {\it Hexagonal circle patterns and integrable systems. Patterns
748: with constant angles,} preprint
749: http://xxx.lanl.gov/abs/math.CV/0109018
750:
751: \bibitem{BPdis} A.~Bobenko,
752: U.~Pinkall, {\it Discrete isothermic surfaces,} J. reine angew.
753: Math. {\bf 475} (1996), 187-208.
754:
755: \bibitem{BPD}
756: A.~Bobenko, U.~Pinkall, {\it Discretization of surfaces and
757: integrable systems,}
758: In: Discrete
759: Integrable Geometry and Physics; Eds. A.I.Bobenko and R.Seiler,
760: Oxford University Press, 1999, 3-58.
761:
762: \bibitem{Doy}
763: K.~Callahan, B.~Rodin, {\it Circle packing immersions form
764: regularly exhaustible surfaces,} Complex Variables {\bf 21}
765: (1993), 171-177.
766:
767: \bibitem{DS} T.Dubejko,
768: K.Stephenson, {\it Circle packings: Experiments in discrete
769: analytic
770: function theory,} Experimental Math., {\bf 4:4}, (1995), 307-348
771:
772:
773: \bibitem{FIK} A.S.~Fokas,
774: A.R.~Its, A.V.~Kitaev, {\it Discrete Painlev\'e equations and
775: their appearance in quantum gravity,} Commun. Math. Phys. {\bf
776: 142} (1991), 313-344.
777:
778: \bibitem{HS}
779: Z.-X.~He, O.~Schramm, {\it The $C^{\infty}$ convergence of
780: hexagonal disc packings to Riemann map,} Acta. Math. {\bf 180}
781: (1998), 219-245.
782:
783: \bibitem{H} Z.-X.~He, {\it
784: Rigidity of infinite disk patterns,} Annals of Mathematics {\bf
785: 149} (1999), 1-33.
786:
787: \bibitem{TH} T.~Hoffman, {\it Discrete CMC surfaces and
788:
789: discrete holomorphic maps,} In: Discrete Integrable Geometry and
790: Physics, Eds.: A.I.Bobenko and R.Seiler, Oxford University
791: Press, 1999, 97-112.
792:
793: \bibitem{IN} A.R.~Its, V.Y.~Novokshenov, {\it The
794:
795: isomonodromic deformation method in the theory of Painlev\'e
796: equations,} Lecture
797:
798: Notes in Math. {\bf 1191}, Springer, Berlin, 1986.
799:
800: \bibitem{MR}
801: A.Marden, B.Rodin, {\it On Thurston's formulation and proof of
802: Andreev's
803: theorem,} Lect. Notes Math., {\bf 1435} (1990), 103-115
804:
805: \bibitem{NC}
806: F.~Nijhoff, H.~Capel, {\it The discrete Korteveg de Vries
807: Equation,}
808:
809: Acta Appl. Math. {\bf 39} (1995), 133-158.
810:
811: \bibitem{NRGO} F.W.~Nijhoff,
812: A.~Ramani, B.~Grammaticos, Y.~Ohta, {\it On discrete Painlev\'e
813: equations associated with lattice KdV systems and the Painlev\'e
814: VI equation,} solv-int/9812011 (1998).
815:
816:
817: \bibitem{RS}
818: B.~Rodin, D.~Sullivan, {\it The convergence of circle packings to
819: Riemann mapping,} J.Diff.Geometry {\bf 26}
820: (1987), 349-360.
821:
822:
823: \bibitem{Schramm} O.~Schramm, {\it
824:
825: Circle patterns with the combinatorics of the square grid,} Duke
826: Math. J.
827:
828: {\bf 86} (1997), 347-389.
829:
830: \bibitem{T} W.P.~Thurston, {\it The finite
831: Riemann mapping theorem,} Invited address, International Symposium
832: in Celebration of the Proof of the Bieberbach Conjecture, Purdue
833: University (1985).
834:
835: \bibitem{Z} A.~Zabrodin, {\it A survey of Hirota's
836: difference equations,} solv-int/9704001 (1997).
837:
838:
839: \end{thebibliography}
840:
841:
842: \end{document}
843: