math0411159/Lame.tex
1: \documentclass[11pt]{article}
2: \usepackage{amsmath,amssymb,theorem}
3: \usepackage[dvips]{graphicx}
4: 
5: \newtheorem{theorem}{Theorem}[section]
6: \newtheorem{proposition}[theorem]{Proposition}
7: \newtheorem{corollary}[theorem]{Corollary}
8: \newtheorem{lemma}[theorem]{Lemma}
9: \newtheorem{remark}[theorem]{Remark}
10: \newtheorem{definition}[theorem]{Definition}
11: \newtheorem{pr}{Proof.}
12: 
13: \renewcommand{\thepr}{}
14: 
15: \newcommand{\proof}[1]{
16: \begin{pr}
17: #1 \,\fbox{}
18: \end{pr}
19: }
20: 
21: 
22: \begin{document}
23: 
24: \title{Lam\'{e} operators with projective octahedral and icosahedral monodromies}
25: \author{Keiri Nakanishi }
26: \date{November 18, 2004}
27: \maketitle
28: 
29: \begin{abstract}
30: 
31: We show that there exists a Lam\'{e} operator $L_{n}$ with projective octahedral monodromy for each 
32: $n \in \tfrac{1}{2}( \mathbb{N} + \tfrac{1}{2}) \cup \tfrac{1}{3}( \mathbb{N} + \tfrac{1}{2}) $, 
33: and with projective icosahedral monodromy for each 
34: $n \in \tfrac{1}{3}( \mathbb{N} + \tfrac{1}{2}) \cup \tfrac{1}{5}( \mathbb{N} + \tfrac{1}{2}) $.
35: To this end, we construct Grothendieck's dessins d'enfants corresponding 
36: to the Belyi morphisms which pull-back hypergeometric operators into Lam\'{e} operators 
37: $L_{n}$ with the desired monodromies. 
38: 
39: 
40: \end{abstract}
41: 
42: \section{Introduction}
43: In this paper, we consider Lam\'{e} differential operators
44: \[ L_{n} = \left(\frac{d}{d x} \right) ^{2} + \frac{1}{2} \left(\frac{1}{x} + \frac{1}{x - 1} + 
45: \frac{1}{x - \lambda}\right)\cdot \frac{d}{d x} - \frac{n(n + 1)x + B}{4x(x - 1)(x - \lambda)} \]
46: whose solutions are algebraic over $\mathbb{C}(x)$.
47: Here, $\lambda$ is a complex number with $\lambda \neq 0, 1$, 
48: $n$ (so-called degree parameter) is a rational number, 
49: and $B \in \mathbb{C}$ is the accessory parameter.
50: The possible finite projective monodromies of $L_{n}$ were studied by Baldassarri, Chiarellotto, 
51: and Dwork, and recently by Beukers, Dahmen, Li\c{t}canu, van der Waall, and Zapponi.
52: One of the most remarkable results is that there are at most finitely many equivalence classes 
53: of $L_{n}$ for fixed $n \in \mathbb{Q}$ and fixed finite monodromy group.
54: This was first done by Chiarellotto~\cite{C}, and later shown by Li\c{t}canu by using the 
55: notion of Grothendieck's dessins d'enfants~\cite{L1}. 
56: For details of the theory of Grothendieck's dessins d'enfants, see~\cite{S} and~\cite{SV}. 
57: Moreover, Chiarellotto and Li\c{t}canu got the explicit formula for the number of equivalence 
58: classes with projective dihedral monodromy of order $2N$ for the case $n = 1$, 
59: which has been generalized more recently by Dahmen~\cite{D} for arbitrary $n$. 
60: They translated the counting problem for the number of equivalence classes of Lam\'{e} operators 
61: into that for the number of the dessins compatible with the ramification data of Belyi morphisms 
62: which pull-back hypergeometric operators into the Lam\'{e} operators. 
63: This strategy is based on the Klein's theorem, 
64: which claims that a second order Fuchsian differential operator with finite projective monodromy 
65: is a rational pull-back of hypergeometric operator in the ``\textit{basic Schwarz list}". 
66: 
67: To carry out this program, the method of Grothendieck's dessins d'enfants 
68: by Litcanu and Dahmen provides a powerful tool. 
69: Baldassarri determined the possible finite projective monodromy groups of $L_{n}$~\cite{B2}, 
70: but recently, Li\c{t}canu got the same results and the necessary conditions for $n$ to have 
71: fixed possible finite projective monodromy group by using the notion of Grothendieck's dessins 
72: d'enfants~\cite{L2}. 
73: By~\cite{B2} and~\cite{L2}, the possible finite projective monodormy groups are dihedral group $\mathcal{D}_{2N}$, 
74: octahedral group $\mathcal{S}_{4}$, or icosahedral group $\mathcal{A}_{5}$. 
75: More recently, Li\c{t}canu~\cite{L2} proves the following theorem:
76: 
77: \vspace{1ex}
78: \noindent
79: \begin{theorem}[~\cite{B2}, ~\cite{L2} Theorem 3.4]\label{Litcanu}
80: 
81: \hspace{2mm}
82: $(1)$ If the projective monodromy group of Lam\'{e} operator $L_{n}$ is dihedral,
83: then $n \in \mathbb{Z}$
84: 
85: \noindent
86: $(2)$ If the projective monodromy group of Lam\'{e} operator $L_{n}$ is octahedral,
87: then $n \in \tfrac{1}{2}( \mathbb{Z} + \tfrac{1}{2}) \cup \tfrac{1}{3}( \mathbb{Z} + 
88: \tfrac{1}{2}) $.
89: 
90: \noindent
91: $(3)$ If the projective monodromy group of Lam\'{e} operator $L_{n}$ is icosahedral,
92: then $n \in \tfrac{1}{3}( \mathbb{Z} + \tfrac{1}{2}) \cup \tfrac{1}{5}( \mathbb{Z} + 
93: \tfrac{1}{2}) $.
94: 
95: \noindent
96: $(4)$ There is no Lam\'{e} operator with projective cyclic monodromy.
97: 
98: \noindent
99: $(5)$ There is no Lam\'{e} operator with projective tetrahedral monodromy.
100: \end{theorem}
101: 
102: \vspace{1ex}
103: The proof of this theorem is based on the analysis of the Belyi morphism which pull-backs 
104: the hypergeometric operator into the Lam\'{e} operator, as well as the combinatorial 
105: data of the corresponding dessin. 
106: Conversely, the following problem arises: 
107: 
108: \vspace{1ex}
109: \noindent
110: \underline{\textbf{Problem}}
111: 
112: \noindent
113: (1) \textit{For each} $n \in \mathbb{Z}$, 
114: \textit{does there exist Lam\'{e} operator} $L_{n}$ 
115: \textit{with projective dihedral monodromy?}
116: 
117: \noindent
118: (2) \textit{For each} $n \in \tfrac{1}{2}(\mathbb{Z} + \tfrac{1}{2}) \cup 
119: \tfrac{1}{3}(\mathbb{Z} + \tfrac{1}{2})$, 
120: \textit{does there exist Lam\'{e} operator} $L_{n}$ 
121: \textit{with projective octahedral monodromy?}
122: 
123: \noindent
124: (3) \textit{For each} $n \in \tfrac{1}{3}(\mathbb{Z} + \tfrac{1}{2}) \cup 
125: \tfrac{1}{5}(\mathbb{Z} + \tfrac{1}{2})$, 
126: \textit{does there exist Lam\'{e} operator} $L_{n}$ 
127: \textit{with projective icosahedral monodromy?}
128: 
129: 
130: \vspace{1ex}
131: If we replace $n$ by $- n - 1$, it is easy to see that $L_{n} = L_{- n - 1}$. 
132: Hence we can assume $n > - \tfrac{1}{2}$. 
133: As we saw, (1) is solved by Beukers and van der Waall~\cite[Theorem5.1]{BW} and Dahmen~\cite{D}. 
134: Beukers and van der Waall~\cite[Theorem 6.1]{BW} and Baldassarri~\cite[(3.e)]{B2} 
135: gave some examples of Lam\'{e} operators which have projective 
136: octahedral monodormy and icosahedral monodromy. 
137: 
138: \vspace{1ex}
139: The aim of this article is to solve (2) and (3) of the problem above. 
140: Assuming $n>-\tfrac{1}{2}$, we have a few possible negative $n$ in each case.
141: Such ``exceptional'' cases will be dealt with case by case in Remarks \ref{rem-exc1} and \ref{rem-exc2}, where we will see that these cases can be easily dismissed. 
142: Thus we may assume $n\geq 0$.
143: In this situation, the following theorem gives the existence of Lam\'e operators $L_n$ with projective octahedral monodromy and projective icosahedral monodromy for each $n$ as in Theorem \ref{Litcanu} (2) and (3):
144: 
145: \vspace{1ex} 
146: \underline{\textbf{Main Theorem}}
147: 
148: \noindent
149: (1) \textit{For each} $n \in \tfrac{1}{2}(\mathbb{N} + \tfrac{1}{2}) 
150: \cup \tfrac{1}{3}(\mathbb{N} + \tfrac{1}{2})$, 
151: \textit{there exists a Lam\'{e} operator} $L_{n}$ \textit{with projective octahedral monodromy.}
152: 
153: \noindent
154: (2) \textit{For each} $n \in \tfrac{1}{3}(\mathbb{N} + \tfrac{1}{2}) 
155: \cup \tfrac{1}{5}(\mathbb{N} + \tfrac{1}{2})$, 
156: \textit{there exists a Lam\'{e} operator} $L_{n}$ \textit{with projective icosahedral monodromy.}
157: 
158: 
159: \vspace{1ex}
160: We will prove this theorem by constructing explicitly the dessins compatible with the ramification data of 
161: the Belyi morphisms which pull-back hypergeometric operators with the same projective 
162: monodromy group into the Lam\'{e} operators $L_{n}$ for each $n$.  
163: From this theorem, we can see that there exist infinitely Lam\'{e} operators with projective 
164: octahedral monodromy and infinitely many ones with projective icosahedral monodromy, which seems unknown. 
165: Note that this theorem does not answer the counting problem of the numbers of the equivalence 
166: classes of Lam\'{e} operators with projective octahedral and icosahedral monodromies. 
167: 
168: \section{Preliminaries.}
169: 
170: Our first aim in this section is to reduce the existences of the Lam\'{e} operators with 
171: projective octahedral (resp. icosahedral) monodromy to the existences of the 
172: Belyi morphisms which pull-back the hypergeometric operators with projective octahedral 
173: (resp. icosahedral) monodromies into the Lam\'{e} operators. 
174: The second aim is to reduce them to the existences of the corresponding dessins. 
175: 
176: \subsection{Hypergeometric operators and Lam\'{e} operators.}
177: 
178: In this subsection, we review some results on hypergeometric operators, their rational 
179: pull-backs, and Lam\'{e} operators. 
180: 
181: \vspace{1ex}
182: Let us first consider the linear differential operator on $\mathbb{P}^{1}$: 
183: \begin{equation}
184: L = D^{n} + a_{1}(z) \cdot D^{n-1} + \cdots + a_{n-1}(z) \cdot D + a_{n}(z), 
185: \label{eq:L}
186: \end{equation}
187: with $D^{i} = (\tfrac{d}{dz})^{i}$ and 
188: $a_{j}(z)$ rational functions in $\mathbb{C}(z)$ 
189: for $1 \leq i, j \leq n$. 
190: The linear operator (\ref{eq:L}) is said to be $Fuchsian$ if 
191: any point on $\mathbb{P}^{1}$ is regular or regular singular. 
192: 
193: Throughout this paper, we treat projective monodromies of the Fuchsian operators 
194: rather than (full) monodromies of them. 
195: Let us consider the natural projection $P$ : 
196: \[
197: P : GL(n, \mathbb{C}) \longrightarrow PGL(n, \mathbb{C}). 
198: \]
199: The monodromy group $G$ of the operator (\ref{eq:L}) is defined in $GL(n, \mathbb{C})$. 
200: Then its natural image of $G$ by $P$ is said to be the \textit{projective monodromy} of operator (\ref{eq:L}), 
201: and we denote it by $PG$, 
202: i.e., $PG = G \cdot Z / Z$ 
203: where $Z = \{\lambda \cdot I_{n} | \lambda \in \mathbb{C}^{\ast} \}$. 
204: Here, $PG$ is a subgroup of $PGL(n, \mathbb{C})$ and determined up to conjugate. 
205: 
206: \vspace{1ex}
207: Let us consider a second order Fuchsian operator with finite projective monodromy. 
208: If it has precisely three regular singular points, it is 
209: the so-called hypergeomtric operator and it has the following normalized form
210: \[
211: H_{ \lambda, \mu, \nu} = \left(\frac{d}{dx} \right)^{2} + \left\{ \frac{1 - \lambda ^{2}}{4x^{2}} 
212: + \frac{1 - \mu ^{2}}{4(x - 1)^{2}} + \frac{\lambda ^{2} + \mu ^{2} + \nu ^{2} - 1}{4x(x - 1)} \right \}, 
213: \]
214: where $\lambda + \mu + \nu > 1$. 
215: The regular singular points of $H_{\lambda, \mu, \nu}$ are $0, 1, \infty$, and 
216: their exponent differences are $\lambda, \mu, \nu,$ respectively. 
217: The finite projective monodromy groups of $H_{\lambda, \mu, \nu}$ 
218: are classified as in the following ``\textit{basic Schwarz list}". 
219: 
220: \begin{center}
221:   \begin{tabular}{|c|c|}                                                                                \hline
222:     \multicolumn{1}{|c|}{$(\lambda, \mu, \nu )$}      & projective monodromy of $H_{\lambda, \mu, \nu}$ \\ \hline 
223:      $(1/n, 1,1/n)$                &         $\mathcal{C}_{n}$ : cyclic of order $n$ \\ \hline
224:      $(1/2, 1/n, 1/2)$     &     $\mathcal{D}_{2n}$ : dihedral of order $2n$ \\ \hline
225:      $(1/2, 1/3, 1/3)$     &                 $\mathcal{A}_{4}$ : tetrahedral \\ \hline
226:      $(1/2, 1/3, 1/4)$     &                  $\mathcal{S}_{4}$ : octahedral \\ \hline
227:      $(1/2, 1/3, 1/5)$     &                 $\mathcal{A}_{5}$ : icosahedral \\ \hline
228:   \end{tabular}
229: \end{center}
230: 
231: 
232: In general, second order Fuchsian operators with finite projective monodromy are characterized by 
233: the following theorem by Klein. 
234: 
235: \begin{theorem}[Klein]
236: Let $L$ be a second order Fuchsian operator with finite projective monodromy $PG$ in normalized form on 
237: $\mathbb{P}^{1}$, i.e. $L = (\tfrac{d}{dx})^{2} + Q(x)$, $Q(x) \in \mathbb{C}(x)$.
238: Then there exists a morphism $f : \mathbb{P}^{1} \rightarrow \mathbb{P}^{1}$ 
239: which ramifies at most over the set $\{0, 1,\infty \}$, 
240: and a unique hypergeometric operator $H$ in the Schwarz list, having the same projective monodromy $PG$, 
241: such that $f^{\ast} H = L$. Moreover, the morphism $f$ as above is unique up to M\"{obius} transformations 
242: except in the case 
243: $(\lambda, \mu, \nu) = (1/2, 1/2, 1/2)$.    
244: \end{theorem}
245: 
246: \proof{
247: \rm{
248: See~\cite{K} or~\cite[Theorem 1.8]{B1}.
249: }
250: }
251: 
252: Let $C$ be an algebraic curve defined over $\mathbb{C}$.
253: A morphism $f : C \rightarrow \mathbb{P}^{1}$ is said to be  \textit{Belyi morphism} 
254: if $f$ has at most three critical values. 
255: 
256: \vspace{1ex}
257: \textit{A Lam\'{e} operator} is a second order Fuchsian operator 
258: having four regular singular points on $\mathbb{P}^{1}$ 
259: with exponent differences $\tfrac{1}{2}, \tfrac{1}{2}, \tfrac{1}{2},$ and $|n + \tfrac{1}{2}|$ 
260: where $n$ is a rational number. 
261: If its four regular singular points are $0, 1, \lambda,$ and $\infty$, 
262: and their exponent differences are $\tfrac{1}{2}, \tfrac{1}{2}, \tfrac{1}{2},$ 
263: and $|n + \tfrac{1}{2}|$ respectively, 
264: then, after suitable transformation, we can assume it has the following Riemann scheme : 
265: 
266: \[
267:   \begin{Bmatrix} 
268:      x = 0 & x = 1 & x = \lambda & x = \infty \\ 
269:      0 & 0 & 0 & - \tfrac{n}{2} \\ 
270:      \tfrac{1}{2} & \tfrac{1}{2} & \tfrac{1}{2} & \tfrac{n+1}{2}  
271:   \end{Bmatrix}, 
272: \]
273: 
274: \vspace{1ex}
275: \noindent
276: and the Lam\'{e} operator $L_{n}$ has the following form: 
277: 
278: \[ 
279: L_{n} = \left(\frac{d}{d x} \right) ^{2} + \frac{1}{2} \left(\frac{1}{x} + \frac{1}{x - 1} + 
280: \frac{1}{x - \lambda}\right)\cdot \frac{d}{dx} - \frac{n(n + 1)x + B}{4x(x - 1)(x - \lambda)}, 
281: \]
282: where $B \in \mathbb{C}$ is the accessory parameter. 
283: Therefore, if there exists a Belyi morphism $f : \mathbb{P}^{1} \rightarrow \mathbb{P}^{1}$ satisfying 
284: the following condition ($\bigstar$) bellow, then $f^{\ast}H_{\lambda, \mu, \nu}$ is a Lam\'{e} operator. 
285: 
286: \vspace{1ex}
287: \textbf{Condition} ($\bigstar$) : 
288: $f^{\ast}H_{\lambda, \mu, \nu}$ has four regular singular points $0, 1, \lambda,$ and $\infty$ 
289: and their exponent differences are 
290: $\tfrac{1}{2}, \tfrac{1}{2}, \tfrac{1}{2},$ and $|n + \tfrac{1}{2}|$ respectively. 
291: 
292: 
293: \vspace{1ex}
294: 
295: We summarize some facts about pull-backs of Fuchsian operators. 
296: We refer to~\cite[Ch.2]{vdW} for details. 
297: 
298: \begin{proposition} \label{prop1}
299: Let $L$ be a Fuchsian operator on $\mathbb{P}^{1}$, 
300: and $f : \mathbb{P}^{1} \rightarrow \mathbb{P}^{1}$ be a morphism. 
301: Then $f^{\ast} L$ is again Fuchsian.
302: \end{proposition}
303: 
304: \proof{
305: \rm{
306: See~\cite[Proposition 2.6.3]{vdW}.
307: }
308: }
309: 
310: \begin{proposition}\label{prop2}
311: Let $L$ be a Fuchsian operator on $\mathbb{P}^{1}$ with projective monodromy $PG_{L}$, 
312: and $f^{\ast}L$ is a pull-back by a morphism 
313: $f : \mathbb{P}^{1} \rightarrow \mathbb{P}^{1}$, with projective monodrory $PG_{f^{\ast}L}$ . 
314: Then $PG_{f^{\ast}L}$ is conjugate in $PGL(2, \mathbb{C})$ to a subgroup of $PG$. 
315: \end{proposition}
316: 
317: \proof{
318: \rm{
319: See~\cite[Corollary 2.6.10]{vdW}.
320: }
321: }
322: 
323: \begin{corollary}\label{cor3} 
324: Let $f : \mathbb{P}^{1} \rightarrow \mathbb{P}^{1}$ be a Belyi morphism satisfying the condition 
325: $($$\bigstar$$)$, 
326: $H_{1/2, 1/3, 1/4}$ $($resp. $H_{1/2, 1/3, 1/5}$$)$ 
327: the hypergeometric operator with projective octahedral $($resp. icosahedral$)$ monodromy. 
328: Let 
329: $n \in \tfrac{1}{2}(\mathbb{Z} + \tfrac{1}{2}) 
330: \cup \tfrac{1}{3}(\mathbb{Z} + \tfrac{1}{2}) 
331: \cup \tfrac{1}{5}(\mathbb{Z} + \tfrac{1}{2})$. 
332: Then the second order Fuchsian operator $f^{\ast}H_{1/2, 1/3, 1/4}$ 
333: $($resp. $f^{\ast}H_{1/2, 1/3, 1/5}$$)$ have projective octahedral 
334: $($resp. icosahedral$)$ monodromy. 
335: \end{corollary}
336: 
337: \proof{
338: \rm{
339: By Theorem \ref{Litcanu}, the projective monodromy of $f^{\ast}H_{1/2, 1/3, 1/4}$ 
340: is octahedral or icosahedral, and by Proposition \ref{prop2}, 
341: it must be conjugate to a subgroup of projective octahedral group. 
342: But projective icosahedral group cannot be conjugate to a subgroup of projective 
343: octahedral group, then the projective monodromy of $f^{\ast}H_{1/2, 1/3, 1/4}$ 
344: is octahedral. The icosahedral case is proved similarly. 
345: }
346: }
347: 
348: By this Corollary, the construction of Lam\'{e} operator with projective octahedral
349: monodromy amounts to the construction of the Belyi morphism satisfying the condition ($\bigstar$). 
350: 
351: 
352: \subsection{Belyi morphisms and Grothendieck's dessins d'enfants.}
353: 
354: This subsection gives some reviews about \textit{Grothendieck's dessins d'enfants}. 
355: For more details, we refer to~\cite{S} and~\cite{SV}. 
356: Let us first recall Belyi's Theorem. 
357: 
358: \begin{theorem}[Belyi's Theorem]
359: Let $X$ be an algebraic curve over $\mathbb{C}$. 
360: Then $X$ is defined over $\bar{\mathbb{Q}}$ if and only if there exist a morphism 
361: $\beta : X \rightarrow \mathbb{P}^{1}(\mathbb{C})$ which ramifies at most over $\{0, 1, \infty \}$.  
362: \end{theorem}
363: 
364: \proof{
365: \rm{
366: Well-known; see~\cite{Be} or~\cite[Theorem I.2]{S}. 
367: }
368: }
369: 
370: Let $\beta : X \rightarrow \mathbb{P}^{1}$ be a Belyi morphism. 
371: For a point $P \in X$, we denote by $e_{P}$ the ramification index at $P$ of $\beta$. 
372: The Belyi morphism $\beta$ is said to be \textit{clean} 
373: if $e_{P} = 2$ for any $P \in \beta^{-1}(1)$, 
374: and \textit{preclean} if $e_{P} \leq 2$ for any $P \in \beta^{-1}(1)$. 
375: Consider a pair $(X, \beta)$ consisting of a complex algebraic curve 
376: defined over $\bar{\mathbb{Q}}$ 
377: and a morphism $\beta : X \rightarrow \mathbb{P}^{1}$. 
378: The pair $(X, \beta)$ is said to be a \textit{Belyi pair} 
379: if the morphism $\beta$ ramifies at most over $\{0, 1, \infty \}$. 
380: Two pairs $(X, \beta)$ and $(Y, \alpha)$ are said to be \textit{isomorphic} 
381: if there exists an isomorphism $\phi : X \rightarrow Y$ 
382: such that $\beta = \alpha \circ \phi$. 
383: 
384: 
385: \begin{definition}[Dessins d'enfants]
386: \rm{
387: Let $X$ be a compact Riemann surface, 
388: $X_{1}$ a connected 1-complex, 
389: $X_{0}$ the set of vertices of $X_{1}$, 
390: $[\iota]$ an isotopical class of inclusions $\iota : X_{1} \hookrightarrow X$. 
391: The triple $D = ( X_{0} \subset X_{1}, [\iota] )$ is said to be 
392: \textit{Grothendieck's dessin d'enfant} on $X$ 
393: if $D$ satisfies the following conditions: 
394: 
395: \noindent 
396: (1) The complement of $X_{0}$ in $X_{1}$ is a finite disjoint union of segments 
397: and each segment is homeomorphic to the interval $(0, 1)$. 
398: 
399: \noindent
400: (2) The complement of $\iota (X_{1})$ in $X$ is a finite disjoint union of open cells
401: (simply connected regions). 
402: 
403: \noindent
404: (3) Each element of $X_{0}$ is equipped with the mark $``\bullet"$ or $``\ast"$ 
405: and if two different elements of $X_{0}$ are connected by a segment, 
406: one is equipped with $``\bullet"$ and another $``\ast"$. 
407: 
408: }
409: \end{definition}
410: 
411: 
412: \begin{definition}
413: \rm{
414: Two Grothendieck's dessins $D = (X_{0} \subset X_{1}, [\iota])$ on $X$ 
415: and $D' = (X'_{0} \subset X'_{1},[\iota '])$ on $X'$ 
416: are said to be \textit{equivalent} if there exists a homeomorphism 
417: $\phi : X \rightarrow X'$ such that 
418: $\phi |_{\iota(X_{1})} : \iota(X_{1}) \rightarrow \iota'(X'_{1})$ and 
419: $\phi |_{\iota(X_{0})} : \iota(X_{0}) \rightarrow \iota'(X'_{0})$ are homeomorphisms. 
420: }
421: \end{definition}
422: 
423: \begin{definition}
424: \rm{
425: A Grothendieck's dessin $D = (X_{0} \subset X_{1},[\iota] )$ 
426: is said to be \textit{preclean} if all vertices with the mark $``\ast"$ 
427: have valencies $\leq 2$. 
428: If all vertices with the mark $``\ast"$ have valencies 2, 
429: $D$ is said to be \textit{clean}. 
430: }
431: \end{definition}
432: 
433: \vspace{1ex}
434: Let $(X, \beta)$ be a Belyi pair. 
435: Then from the Belyi pair $(X, \beta)$, we can construct a dessin 
436: $D = (\beta^{-1}(\{0, 1 \}) \subset \beta^{-1}([0, 1]))$ 
437: by putting the mark $``\bullet"$ on the vertices of $\beta^{-1}(0)$, 
438: and $``\ast"$ on the vertices of $\beta^{-1}(1)$. 
439: 
440: 
441: \begin{theorem}[Grothendieck Correspondence]
442: This correspondence gives a bijection between 
443: the set of isomorphic classes of preclean Belyi pairs
444: and the set of equivalence classes of preclean dessins. 
445: \end{theorem}
446: 
447: \proof{
448: \rm{
449: See~\cite[Theorem I.5]{S}. 
450: }
451: }
452: 
453: The procedure for getting Belyi pairs from dessins is given in~\cite[Chapter I, \S3]{S}. 
454: By this correspondence, the most important thing is that 
455: the ramification multiplicities of points in $\beta^{-1}(0)$ 
456: (resp. $\beta^{-1}(1)$) 
457: are translated to the valencies of $``\bullet"$ (resp. $``\ast"$). 
458: 
459: \vspace{1ex}
460: By Grothendieck Correspondence, we can construct Lam\'{e} operator $L_{n}$ with projective 
461: octahedral and icosahedral monodromy, if there exists a dessin d'enfant corresponding to a Belyi 
462: morphism which satisfies the condition ($\bigstar$). 
463: In the next section, we are going to construct the dessins that the corresponding Belyi 
464: morphisms satisfies the condition ($\bigstar$) for each $n$.
465: 
466: \begin{definition}
467: 
468: \rm{
469: A Belyi morphism $f : \mathbb{P}^{1} \rightarrow \mathbb{P}^{1}$ 
470: is said to be \textit{ $\ast$-morphism } 
471: if $\{0, 1,\infty \} \subseteq f^{-1} ( \{0, 1,\infty \})$.
472: }
473: \end{definition}
474: 
475: 
476: \begin{remark}
477: \rm{
478: Under the action of $PGL(2, \mathbb{C})$, any Belyi morphism 
479: $f : \mathbb{P}^{1} \rightarrow \mathbb{P}^{1}$ 
480: is transformed to a $\ast$-morphism. 
481: }
482: \end{remark}
483: 
484:  
485: 
486: \section{Constructions of the Dessins.}
487: 
488: \subsection{The case of projective octahedral monodromy.} 
489: 
490: We start this subsection by preparing some notations. 
491: We denote a second order Fuchsian differential operator on $\mathbb{P}^{1}$ by $L$ 
492: and its exponent difference at $P \in \mathbb{P}^{1}$ by $\Delta _{P, L}$. 
493: Set  
494: $\Delta _{L} = \displaystyle \sum_{P \in \mathbb{P}^{1}} ( \Delta_{P, L} - 1 ) $. 
495: 
496: \vspace{1ex} 
497: We need the following useful lemma. 
498: 
499: \begin{lemma}[~\cite{BD} Lemma 1.5]\label{Bal-Dw}
500: Let $f : \mathbb{P}^{1} \rightarrow \mathbb{P}^{1}$ be a morphism, 
501: and $L$ a Fuchsian second order differential operator. 
502: Then 
503: \[ 
504: \mathrm{deg}(f) = \frac{\Delta_{f^{\ast} L} + 2}{\Delta_{L} + 2}. 
505: \] 
506: \end{lemma}
507: 
508: \proof{
509: \rm{
510: This is the genus 0 case of~\cite[Lemma 1.5]{BD}. 
511: Let $Q \in \mathbb{P}^{1}$, $P \in \mathbb{P}^{1}$ be points with $f(Q) = P$. 
512: If $\alpha_{1}, \alpha_{2}$ are local exponents of $L$ at $P$, 
513: then the local exponents of $f^{\ast}L$ at $Q$ are $\alpha_{1} \cdot e_{Q, f}$ and 
514: $\alpha_{2} \cdot e_{Q, f}$ where $e_{Q, f}$ is the ramification index of $f$ at $Q$. 
515: Thus we get 
516: \[
517: \Delta_{Q, f^{\ast}L} = \Delta_{P, L} \cdot e_{Q, f}, 
518: \]
519: whence having 
520: \[
521: \sum_{Q \mapsto P} \Delta_{Q, f^{\ast}L} = \mathrm{deg}(f) \cdot \Delta_{P, L}. 
522: \]
523: 
524: Now let $S$ be finite subset of $\mathbb{P}^{1}$ and put 
525: \[
526: \Delta (L, S) = \sum_{P \in S} (\Delta_{P, L} - 1) . 
527: \]
528: 
529: \noindent
530: Here we have 
531: 
532: \begin{equation}
533: \Delta (f^{\ast}L, f^{-1}(S)) + \# f^{-1}(S) 
534: = \mathrm{deg}(f) \cdot \{ \Delta(L, S) + \# S \}, 
535: \label{eq:3-1-1}
536: \end{equation}
537: 
538: \noindent
539: and by Riemann-Hurwitz formula, 
540: 
541: \begin{equation}
542: -2 + 2 \cdot \mathrm{deg}(f) = \mathrm{deg}(f) \cdot (\# S) - \# f^{-1}(S). 
543: \label{eq:3-1-2}
544: \end{equation}
545: 
546: \noindent
547: When we take $\# S$ sufficiently large, (\ref{eq:3-1-1}) and (\ref{eq:3-1-2}) imply 
548: \[
549: \mathrm{deg}(f) \cdot (\Delta_{L} + 2) = \Delta_{f^{\ast}L} + 2, 
550: \]
551: and the lemma follows. 
552: }
553: }
554:  
555: 
556: \vspace{1ex}
557: Let $L_{n}$ denote a Lame operator with projective octahedral monodromy. 
558: By Theorem \ref{Litcanu}, we have $n \in \tfrac{1}{2}(\mathbb{N} + \tfrac{1}{2}) 
559: \cup \tfrac{1}{3}(\mathbb{N} + \tfrac{1}{2})$. 
560: As we saw in \S2.1, there is a $\ast$-morphism such that 
561: $L_{n} = f^{\ast}H_{1/2, 1/3, 1/4}$. 
562: We want to construct such a $\ast$-morphism $f : \mathbb{P}^{1} \rightarrow \mathbb{P}^{1}$. 
563: 
564: 
565: In the octahedral case, by Lemma \ref{Bal-Dw}, 
566: \[
567: \mathrm{deg}(f) = 12n, 
568: \] 
569: and Riemann-Hurwitz formula implies 
570: \[
571: \# f^{-1} (\{0, 1, \infty \}) = 12 n + 2.
572: \] 
573: Then we can assume 
574: $f^{-1}(\{0, 1, \infty \}) = \{0, 1, \lambda, \infty, a_{1}, \cdots a_{12n-2} \}$ 
575: where, $a_{1}, \cdots a_{12n-2}$ denote distinct points different from $0, 1, \lambda, \infty$, 
576: and thus, possible ramification data of such an $f$ is given as follows: 
577: 
578: \begin{center}
579:   \begin{tabular}{|c|c|c|c|c|c|c|}                                                                       \hline
580:      {}      & 0    & 1    & $\lambda$ & $\infty$               & $a_{1}, \cdots a_{12n-2}$ & deg     \\ \hline 
581:      0       & 0, 1 & 0, 1 & 0, 1      & 0, $2n+1$              & 0, 2                      & $12n$   \\ \hline
582:      1       & 0    & 0    & 0         & 0, $3n+ \tfrac{3}{2}$  & 0, 3                      & $12n$   \\ \hline
583:      $\infty$& 0, 2 & 0, 2 & 0, 2      & 0, $4n+2$              & 0, 4                      & $12n$   \\ \hline
584:   \end{tabular}
585: \end{center}
586: 
587: \vspace{1ex}
588: Here, we explain how to read this table. 
589: For $P \in \{0, 1, \infty \}$ (an entry of the first column) and 
590: $Q \in \{0, 1, \infty, a_{1}, \cdots a_{12n-2} \}$ (an entry of the first row), 
591: the possible ramification index of $f$ at $P$ is written in the corresponding entry 
592: (i.e., $(Q, P)$-th entry); the number 0 occurs when $f(Q) \neq P$. 
593: These values $e_{Q, P}$ are calculated by the formula 
594: \[
595: \Delta_{Q, L_{n}} = e_{Q, f} \cdot \Delta_{P, H} 
596: \]
597: where $H = H_{1/2, 1/3, 1/4}$. 
598: Moreover, these values must satisfy the following compatibility conditions:
599: 
600: 
601: $(1)$ The summation of every row is equal to deg$(f)$. 
602: 
603: $(2)$ Every column contains only one non-zero number. 
604: 
605: \vspace{1ex}
606: Let us ask, conversely, if we can construct the $\ast$-morphism $f$, 
607: or what amounts to the same, the corresponding dessin, starting from a table 
608: as above which satisfies the above compatibility conditions for each
609: $n \in \tfrac{1}{2}( \mathbb{N} + \tfrac{1}{2}) \cup \tfrac{1}{3}( \mathbb{N} + \tfrac{1}{2})$. 
610: 
611: 
612: 
613: \vspace{1ex}
614: \underline{\textbf{(1) The case for $n \in \tfrac{1}{2}( \mathbb{N} + \tfrac{1}{2}$)}}
615: 
616: \begin{center}
617:   \begin{tabular}{|c|c|c|c|c|c|c|}                                                                                       \hline
618:      {}      & 0 & 1 & $\lambda$ & $\infty$ & $a_{1}, \cdots a_{12n-2}$                 & deg     \\ \hline 
619:      0       & 1 & 1 & 1         & 0        & $( 6n-\tfrac{3}{2})$ pts with $mult. = 2$ & $12n$   \\ \hline
620:      1       & 0 & 0 & 0         & 0        & $4n$ pts with $mult. = 3$                 & $12n$   \\ \hline
621:      $\infty$& 0 & 0 & 0         & $4n+2$   & $( 2n-\tfrac{1}{2})$ pts with $mult. = 4$ & $12n$   \\ \hline
622:   \end{tabular}
623: \end{center}
624: 
625: 
626: If the dessin has $N$ loops with valency 4, 
627: \[
628: N = 2n - \tfrac{1}{2} \Leftrightarrow n = \tfrac{1}{2}(N + \tfrac{1}{2}) \in 
629: \tfrac{1}{2}(\mathbb{N} + \tfrac{1}{2}) 
630: \] 
631: and then the table becomes the following. 
632: 
633: \begin{center}
634:   \begin{tabular}{|c|c|c|c|c|c|c|}                                                                                       \hline
635:      {}      & 0 & 1 & $\lambda$ & $\infty$ & $a_{1}, \cdots a_{12n-2}$         &   deg        \\ \hline 
636:      0       & 1 & 1 & 1         & 0        & $3N$ pts with $mult. = 2$         & $ 6N + 3 $   \\ \hline
637:      1       & 0 & 0 & 0         & 0        & $( 2N + 1 )$ pts with $mult. = 3$ & $ 6N + 3 $   \\ \hline
638:      $\infty$& 0 & 0 & 0         & $2N+3$   & $  N  $ pts with $mult. = 4$      & $ 6N + 3 $   \\ \hline
639:   \end{tabular}
640: \end{center}
641: So it suffices to construct the dessins compatible with the table 
642: for all $N \in \mathbb{N}$. 
643: Now, we construct dessins.
644: 
645: \vspace{1ex} 
646: For $N = 0$,  
647: \begin{center}
648: \includegraphics[width=10em,clip]{1-0.eps}
649: \end{center}
650: 
651: For $N=1$, 
652: \begin{center}
653: \includegraphics[width=12em,clip]{1-1.eps}
654: \end{center}
655: 
656: For $N=2$,  
657: \begin{center}
658: \includegraphics[width=12em,clip]{1-2.eps}
659: \end{center}
660: 
661: For $N=k \geq 2$, 
662: \begin{center}
663: \includegraphics[width=15em,clip]{1-3.eps}
664: \end{center}
665: 
666: It is easy to check that these are the dessins we want, and thus we could draw dessins inductively for all $N \in \mathbb{N}$. 
667: 
668: \begin{remark}
669: \rm{
670: For each $N \geq 1$, our dessin drawn above is one of those which are 
671: compatible with the table. 
672: There may exist other dessins compatible with the table. 
673: }
674: \end{remark}
675: 
676: \underline{\textbf{(2) The case for $n \in \tfrac{1}{3}( \mathbb{N} + \tfrac{1}{2}$)}} 
677: 
678: \begin{center}
679:   \begin{tabular}{|c|c|c|c|c|c|c|}                                                                            \hline
680:      {}      & 0 & 1 & $\lambda$ & $\infty$          & $a_{1}, \cdots a_{12n-2}$                 & deg     \\ \hline 
681:      0       & 1 & 1 & 0         & 0                 & $( 6n-1 )$ pts with $mult. = 2$           & $12n$   \\ \hline
682:      1       & 0 & 0 & 0         & $3n+\tfrac{3}{2}$ & $( 3n-\tfrac{1}{2})$ pts with $mult. = 3$ & $12n$   \\ \hline
683:      $\infty$& 0 & 0 & 2         & 0                 & $( 3n-\tfrac{1}{2})$ pts with $mult. = 4$ & $12n$   \\ \hline
684:   \end{tabular}
685: \end{center}
686: 
687: If the dessin has $N$ loops with valency 4, 
688: \[
689: N = 3n - \tfrac{1}{2} \Leftrightarrow n = \tfrac{1}{3}(N + \tfrac{1}{2}) \in 
690: \tfrac{1}{3}(\mathbb{N} + \tfrac{1}{2}) 
691: \] 
692: and then the table becomes the following. 
693: 
694: \begin{center}
695:   \begin{tabular}{|c|c|c|c|c|c|c|}                                                                            \hline
696:      {}      & 0 & 1 & $\lambda$ & $\infty$ & $a_{1}, \cdots a_{12n-2}$ & deg      \\ \hline 
697:      0       & 1 & 1 & 0         & 0        & $2N$ pts with $mult. = 2$ & $4N + 2$ \\ \hline
698:      1       & 0 & 0 & 0         & $N + 2$  & $N$ pts with $mult. = 3$  & $4N + 2$ \\ \hline
699:      $\infty$& 0 & 0 & 2         & 0        & $N$ pts with $mult. = 4$  & $4N + 2$ \\ \hline
700:   \end{tabular}
701: \end{center}
702: As in the previous case, it suffices to construct the dessins compatible with the table 
703: for all $N \in \mathbb{N}$. 
704: 
705: Now, we construct dessins. 
706: (For each $N \geq 1$, our dessin is one of those which are compatible with the table.) 
707: 
708: \vspace{1ex}
709: For $N = 0$, 
710: \begin{center}
711: \includegraphics[width=10em,clip]{2-0.eps}
712: \end{center}
713: 
714: For $N=1$, 
715: \begin{center}
716: \includegraphics[width=12em,clip]{2-1.eps}
717: \end{center}
718: 
719: For $N=2$, 
720: \begin{center}
721: \includegraphics[width=12em,clip]{2-2.eps}
722: \end{center}
723: 
724: For $N=3$, 
725: \begin{center}
726: \includegraphics[width=10em,clip]{2-3.eps}
727: \end{center}
728: 
729: For $N=4$, 
730: \begin{center}
731: \includegraphics[width=12em,clip]{2-4.eps}
732: \end{center}
733: 
734: For $N=5$, 
735: \begin{center}
736: \includegraphics[width=12em,clip]{2-5.eps}
737: \end{center}
738: 
739: For $N=6$, 
740: \begin{center}
741: \includegraphics[width=12em,clip]{2-6.eps}
742: \end{center}
743: 
744: 
745: For $N \geq 6$, we can construct dessins inductively according to 
746: the following operations. We operate the lower-half part of the dessin. 
747: (We draw only the part which is enclosed by the dotted line in the dessin of $N=6$.) 
748: 
749: If the dessin with $N$ loops is 
750:  
751: \begin{center}
752: \includegraphics[width=15em,clip]{2-7.eps}
753: \end{center}
754: 
755: \noindent
756: then the dessin with $N+1$ loops is obtained by  
757: 
758: \begin{center}
759: \includegraphics[width=15em,clip]{2-8.eps}
760: \end{center}
761: 
762: \noindent
763: and the dessin with $N+2$ loops is obtained by 
764:  
765: \begin{center}
766: \includegraphics[width=15em,clip]{2-9.eps}
767: \end{center}
768: 
769: We can see that the part which is enclosed by the dotted line 
770: in the above dessin repeatedly appears when $N$ is even. 
771: Also we can easily see that dessins inductively constructed by this operation 
772: are compatible with the table above. 
773: In fact, in each step of this operation, 
774: the number of $``\bullet"$ with valency 2 increase by two, 
775: that of $``\ast"$ with valency 3 increase by one, 
776: that of edges increase by four, and 
777: that of loops having $a_{i}$ in the fiber over $\infty$ with valency 4 inside 
778: increase by one. 
779: 
780: \begin{remark}\label{rem-exc1}
781: \rm{
782: As we saw in the introduction, we can assume degree parameter $n > -1/2$. 
783: If there exists Lam\'{e} operator with projective octahedral monodromy 
784: for $n=-1/4$ (resp. $n=-1/6$), 
785: there would exist Belyi morphism which pull-backs $H_{1/2, 1/3, 1/4}$ 
786: into $L_{-1/4}$ (resp. $L_{-1/6}$). 
787: But this morphism has to have negative degree, and hence the corresponding dessin does not exist. 
788: Therefore, the Lam\'{e} operator with projective octahedral monodromy 
789: for $n=-1/4$ and $n=-1/6$ do not exist. 
790: }
791: \end{remark}
792: 
793: \subsection{The case of projective icosahedral monodromy.}
794: 
795: As in the previous subsection, 
796: We want $\ast$ - morphism $f : \mathbb{P}^{1} \rightarrow \mathbb{P}^{1}$ such that 
797: $L_{n} = f^{\ast}H_{1/2, 1/3, 1/5}$ for 
798: $n \in \tfrac{1}{3}( \mathbb{N} + \tfrac{1}{2}) \cup \tfrac{1}{5}( \mathbb{N} + \tfrac{1}{2}) $. 
799: 
800: Lemma \ref{Bal-Dw} and Riemann-Hurwitz formula implies 
801: \[
802: \# f^{-1} (\{0, 1, \infty \}) = 30 n + 2.
803: \] 
804: So we can assume 
805: $f^{-1}(\{0, 1, \infty \}) = \{0, 1, \lambda, \infty, a_{1}, \cdots a_{30n-2} \}$, 
806: and possible ramification data of $f$ is according to the following table. 
807: 
808: \begin{center}
809:   \begin{tabular}{|c|c|c|c|c|c|c|}                                                                       \hline
810:      {}      & 0    & 1    & $\lambda$ & $\infty$               & $a_{1}, \cdots a_{12n-2}$ & deg     \\ \hline 
811:      0       & 1    & 1    & 1         & 0, $2n+1$              & 0, 2                      & $30n$   \\ \hline
812:      1       & 0    & 0    & 0         & 0, $3n+ \tfrac{3}{2}$  & 0, 3                      & $30n$   \\ \hline
813:      $\infty$& 0    & 0    & 0         & 0, $5n+ \tfrac{5}{2}$  & 0, 5                      & $30n$   \\ \hline
814:   \end{tabular}
815: \end{center}
816: 
817: In this subsection, we construct dessins compatible with the ramification data above for each 
818: $n \in \tfrac{1}{3}( \mathbb{N} + \tfrac{1}{2}) \cup \tfrac{1}{5}( \mathbb{N} + \tfrac{1}{2}) $. 
819: 
820: \vspace{1ex}
821: \underline{\textbf{(3) The case for $n \in \tfrac{1}{3}( \mathbb{N} + \tfrac{1}{2}$)}}
822: 
823: \begin{center}
824:   \begin{tabular}{|c|c|c|c|c|c|c|}                                                                                       \hline
825:      {}      & 0 & 1 & $\lambda$ & $\infty$          & $a_{1}, \cdots a_{12n-2}$                  & deg     \\ \hline 
826:      0       & 1 & 1 & 1         & 0                 & $( 15n-\tfrac{3}{2})$ pts with $mult. = 2$ & $30n$   \\ \hline
827:      1       & 0 & 0 & 0         & $3n+\tfrac{3}{2}$ & $( 9n-\tfrac{1}{2})$ pts with $mult. = 3$  & $30n$   \\ \hline
828:      $\infty$& 0 & 0 & 0         & 0                 & $6n$ pts with $mult. = 5$                  & $30n$   \\ \hline
829:   \end{tabular}
830: \end{center}
831: 
832: When we have $n \in \tfrac{1}{3}(\mathbb{N} + \tfrac{1}{2})$, 
833: $6n$ is odd and we can put $6n = 2M + 1$ where $M \in \mathbb{N}$. 
834: Then the table becomes the following. 
835: 
836: \begin{center}
837:   \begin{tabular}{|c|c|c|c|c|c|c|}                                                                                       \hline
838:      {}      & 0 & 1 & $\lambda$ & $\infty$ & $a_{1}, \cdots a_{12n-2}$         & deg        \\ \hline 
839:      0       & 1 & 1 & 1         & 0        & $( 5M + 1 )$ pts with $mult. = 2$ & $10 M + 5$ \\ \hline
840:      1       & 0 & 0 & 0         & $M + 2$  & $( 3M + 1 )$ pts with $mult. = 3$ & $10 M + 5$ \\ \hline
841:      $\infty$& 0 & 0 & 0         & 0        & $( 2M + 1 )$ pts with $mult. = 5$ & $10 M + 5$ \\ \hline
842:   \end{tabular}
843: \end{center}
844: So it suffices to construct the dessins compatible with the table 
845: for all $M \in \mathbb{N}$. 
846: Now, we construct dessins. 
847: (For each $M \geq 1$, our dessin is one of those which are compatible with the table.) 
848: 
849: \vspace{1ex}
850: For $M=0$, 
851: \begin{center}
852: \includegraphics[width=12em,clip]{3-0.eps}
853: \end{center}
854: 
855: For $M=1$, 
856: \begin{center}
857: \includegraphics[width=12em,clip]{3-1.eps}
858: \end{center}
859: 
860: For $M=2$, 
861: \begin{center}
862: \includegraphics[width=12em,clip]{3-2.eps}
863: \end{center}
864: 
865: For $M=3$, 
866: \begin{center}
867: \includegraphics[width=12em,clip]{3-3.eps}
868: \end{center}
869: 
870: For $M=4$, 
871: \begin{center}
872: \includegraphics[width=12em,clip]{3-4.eps}
873: \end{center}
874: 
875: For $M=5$, 
876: \begin{center}
877: \includegraphics[width=15em,clip]{3-5.eps}
878: \end{center}
879: (We draw only the upper-half part, the lower part looks like the same as $M=4$.) 
880: 
881: For $M=6$, 
882: \begin{center}
883: \includegraphics[width=15em,clip]{3-6.eps}
884: \end{center}
885: (We draw only the main part which is enclosed by the dotted line in the dessin of $M=5$, 
886: the other parts are same as $M=5$.) 
887: 
888: \vspace{1ex}
889: For $M \geq 6$, we can construct dessins inductively according to 
890: the following operations. We operate the upper half part of the dessin. 
891: (We draw only the main part that is enclosed by the dotted line in the dessin of $M=6$.) 
892: 
893: If the dessin with $M=2m$ loops is 
894:  
895: \begin{center}
896: \includegraphics[width=15em,clip]{3-7.eps}
897: \end{center}
898: 
899: \noindent
900: then the dessin with $2m+1$ loops is obtained by  
901: 
902: \begin{center}
903: \includegraphics[width=15em,clip]{3-8.eps}
904: \end{center}
905: 
906: \noindent
907: and the dessin with $2m+2$ loops is obtained by 
908:  
909: \begin{center}
910: \includegraphics[width=15em,clip]{3-9.eps}
911: \end{center}
912: 
913: We can easily see that the part which is enclosed by the dotted line 
914: in the above dessin repeatedly appears when $M$ is even 
915: and that dessins inductively constructed by this operation 
916: are compatible with the table. 
917: 
918: \vspace{2ex}
919: \underline{\textbf{(4) The case for $n \in \tfrac{1}{5}( \mathbb{N} + \tfrac{1}{2}$)}}
920: 
921: \begin{center}
922:   \begin{tabular}{|c|c|c|c|c|c|c|}                                                                                       \hline
923:      {}      & 0 & 1 & $\lambda$ & $\infty$          & $a_{1}, \cdots a_{12n-2}$                  & deg     \\ \hline 
924:      0       & 1 & 1 & 1         & 0                 & $( 15n-\tfrac{3}{2})$ pts with $mult. = 2$ & $30n$   \\ \hline
925:      1       & 0 & 0 & 0         & 0                 & $10n$ pts with $mult. = 3$                 & $30n$   \\ \hline
926:      $\infty$& 0 & 0 & 0         & $5n+\tfrac{5}{2}$ & $( 5n-\tfrac{1}{2})$ pts with $mult. = 5$                  & $30n$   \\ \hline
927:   \end{tabular}
928: \end{center}
929: 
930: If the dessin has $N$ loops with valency 5, 
931: \[
932: N = 5n - \tfrac{1}{2} \Leftrightarrow n = \tfrac{1}{5}(N + \tfrac{1}{2}) \in 
933: \tfrac{1}{5}(\mathbb{N} + \tfrac{1}{2}) 
934: \] 
935: and then the table becomes the following. 
936: 
937: \begin{center}
938:   \begin{tabular}{|c|c|c|c|c|c|c|}                                                                                       \hline
939:      {}      & 0 & 1 & $\lambda$ & $\infty$ & $a_{1}, \cdots a_{12n-2}$       & deg      \\ \hline 
940:      0       & 1 & 1 & 1         & 0        & $ 3N $ pts with $mult. = 2$     & $6N + 3$ \\ \hline
941:      1       & 0 & 0 & 0         & 0        & $(2N + 1)$ pts with $mult. = 3$ & $6N + 3$ \\ \hline
942:      $\infty$& 0 & 0 & 0         & $N + 3$  & $ N $ pts with $mult. = 5$      & $6N + 3$ \\ \hline
943:   \end{tabular}
944: \end{center}
945: So it suffices to construct the dessins compatible with the table 
946: for all $N \in \mathbb{N}$. 
947: Now, we construct dessins. 
948: (For each $N \geq 1$, our dessin is one of those which are compatible with the table.) 
949: 
950: \vspace{1ex}
951: For $N=0$, 
952: \begin{center}
953: \includegraphics[width=10em,clip]{4-0.eps}
954: \end{center}
955: 
956: For $N=1$, 
957: \begin{center}
958: \includegraphics[width=10em,clip]{4-1.eps}
959: \end{center}
960: 
961: For $N=2$, 
962: \begin{center}
963: \includegraphics[width=10em,clip]{4-2.eps}
964: \end{center}
965: 
966: For $N=3$, 
967: \begin{center}
968: \includegraphics[width=10em,clip]{4-3.eps}
969: \end{center}
970: 
971: For $N=4$, 
972: \begin{center}
973: \includegraphics[width=12em,clip]{4-4.eps}
974: \end{center}
975: 
976: For $N=5$, 
977: \begin{center}
978: \includegraphics[width=12em,clip]{4-5.eps}
979: \end{center}
980: 
981: For $N=6$, 
982: \begin{center}
983: \includegraphics[width=15em,clip]{4-6.eps}
984: \end{center}
985: 
986: \vspace{1ex}
987: For $N \geq 6$, we can construct dessins inductively according to 
988: the following operations. We operate the left side of the dessin. 
989: (We draw only the main part which is enclosed by the dotted line in the dessin of $N=6$.) 
990: 
991: If the dessin with $N=2k, (k \geq 3)$ loops is 
992:  
993: \begin{center}
994: \includegraphics[width=15em,clip]{4-7.eps}
995: \end{center}
996: 
997: \noindent
998: then the dessin with $N+1=2k+1$ loops is obtained by  
999: 
1000: \begin{center}
1001: \includegraphics[width=15em,clip]{4-8.eps}
1002: \end{center}
1003: 
1004: \noindent
1005: and the dessin with $N+2=2k+2$ loops is obtained by 
1006:  
1007: \begin{center}
1008: \includegraphics[width=15em,clip]{4-9.eps}
1009: \end{center}
1010: 
1011: We can easily see that the part which is enclosed by the dotted line 
1012: in the above dessin repeatedly appears when $N$ is even 
1013: and that dessins inductively constructed by this operation 
1014: are compatible with the table. 
1015: 
1016: \begin{remark}\label{rem-exc2}
1017: \rm{
1018: The ``exceptional'' values of $n$ in icosahedral case are $n=-1/6$ and $n=-1/10$.
1019: Similarly to the case of octahedral monodromy, we can show that these cases do not occur.
1020: }
1021: \end{remark}
1022: 
1023: \vspace{1ex}
1024: \textbf{Acknowledgements.}
1025: I would like to thank R\u{a}zvan Li\c{t}canu for showing me the new version of~\cite{L2}. 
1026: I am deeply grateful to my adviser, Professor Fumiharu Kato for his advice and support.
1027: Thanks are also due to Masao Aoki and Koki Itoh for their useful comments.
1028: 
1029: \begin{thebibliography}{99}
1030: 
1031: \bibitem[Be]{Be}
1032: G. V. Belyi, Galois extensions of a maximal cyclotomic field, 
1033: \textit{Math. USSR Izv.} \textbf{14} (1980) no.2. 247-256. 
1034: 
1035: \bibitem[BD]{BD}
1036: F. Baldassarri and B. Dwork, On second order differential equations with algebraic solutions,
1037: \textit{Amer.J.Math.} \textbf{101} (1979) 42-76.
1038: 
1039: \bibitem[B1]{B1}
1040: F. Baldassarri, On second order linear differential equations on algebraic curves.
1041: \textit{Amer.J.Math.} \textbf{102} (1980) 517-535.
1042: 
1043: \bibitem[B2]{B2}
1044: F. Baldassarri, On algebraic solutions of Lam\'{e}'s differential equations.
1045: \textit{J.Differential Equations.} \textbf{41} (1981) 44-58.
1046: 
1047: \bibitem[BW]{BW}
1048: F. Beukers and A van der Waall, Lam\'{e} equations with algebraic solutions.
1049: \textit{J.Differential Equations.} \textbf{197} (2004) no.1. 1-25.
1050: 
1051: \bibitem[C]{C}
1052: B. Chiarellotto, On Lam\'{e} operators which are pull backs of hypergeometric ones.
1053: \textit{Trans.Amer.Math.Soc.} \textbf{347} (1995) no.8. 2753-2780.
1054: 
1055: \bibitem[D]{D}
1056: S. Dahmen, Counting Integral Lam\'{e} Equations by Means of Dessins d'Enfants.
1057:  arXiv:math.CA/0311510.
1058: 
1059: \bibitem[K]{K}
1060: F. Klein, Vorlesungen \"{u}ber das Ikosaeder.
1061: B. G. Teubner, Leipzig, 1884.
1062: 
1063: \bibitem[L1]{L1}
1064: R. Li\c{t}canu, Counting Lam\'{e} differential operators.
1065: \textit{Rend. Sem. Mat. Univ. Padova.} \textbf{107} (2002) 191-208.
1066: 
1067: \bibitem[L2]{L2}
1068: R. Li\c{t}canu, Lam\'{e} operators with finite monodromy - a combinatorial approach. 
1069: \textit{J.Differential Equations.} doi:10.1016/j.jde.2004.08.012. 
1070: 
1071: \bibitem[M]{M}
1072: R. S. Maier, Algebraic solotions of the Lam\'{e} equation, revisited.
1073: \textit{J. Differential Equations} \textbf{198} (2004) no.1 16-34.
1074: 
1075: \bibitem[S]{S}
1076: L. Schneps, Dessins d'enfants on the Riemann sphere, 
1077: in L. Schneps (Ed.), ``The Grothendieck theory of dessins d'enfants", 
1078: London Math. Soc. Lecture Note Series 200, Cambridge Univ. Press, 1994.
1079: 
1080: \bibitem[SV]{SV}
1081: G. B. Shabat and V. A. Voevodsky, Drawing Curves over number fields, 
1082: in P. Cartier et al (Eds.), ``Grothendieck Fertschrift III", 
1083: Progress in Math. 88, Birkh\"{a}user, Basel. 1990, 199-227.
1084: 
1085: \bibitem[vdW]{vdW}
1086: A. van der Waall, Lam\'{e} Equations with Finite Monodromy.
1087: \textit{Universiteit Utrecht}, Thesis, 2002. 
1088: 
1089: \bibitem[Z]{Z}
1090: L.Zapponi, Some arithmetic properties of Lam\'{e} operators with dihedral monodromy.
1091:  arXiv:math.AG/0403287.
1092: 
1093: \end{thebibliography}
1094: 
1095: 
1096: \noindent
1097: Department of Mathematics, Faculty of Science, Kyoto University, Kyoto 606-8502, JAPAN.
1098: 
1099: \noindent
1100: E-mail: \texttt{nakanisi@math.kyoto-u.ac.jp}
1101: 
1102: \end{document}