1: \documentclass[12pt]{article}
2:
3: \usepackage{amssymb}
4:
5:
6: \def\qed{{\hfill $\diamondsuit$}}
7: \def\CP{{{\mathbb C}{\rm P}}}
8: \def\C{{\mathbb C}}
9: \def\Aut{{\rm Aut}}
10:
11:
12: \usepackage{theorem}
13:
14: \newtheorem{theorem}{Theorem}
15: \newtheorem{proposition}{Proposition}[section]
16: \newtheorem{corollary}[proposition]{Corollary}
17: \newtheorem{lemma}[proposition]{Lemma}
18: \newtheorem{conjecture}[proposition]{Conjecture}
19: {\theorembodyfont{\rmfamily}
20: \newtheorem{definition}[proposition]{Definition}
21: \newtheorem{example}[proposition]{Example}
22: \newtheorem{remark}[proposition]{Remark}
23: \newtheorem{notation}[proposition]{Notation}
24: }
25:
26: \include{epsf}
27:
28: \title{Enumeration of almost polynomial rational functions
29: with given critical values}
30:
31:
32: \author{Dmitri Panov\thanks{IPDE, IHES, Le Bois-Marie, 35,
33: route de Chartres F-91440, Bures-sur-Yvette, France.
34: E-mail: panov@ihes.fr\,.},
35: Dimitri Zvonkine\thanks{
36: Institut Math{\'e}matique de Jussieu,
37: Universit{\'e} Paris~VI, 175, rue du Chevaleret, 75013 Paris,
38: France.
39: E-mail: zvonkine@math.jussieu.fr\,.\/\/
40: The second author was partially suported by
41: EAGER - European Algebraic Geometry Research Training Network,
42: contract No. HPRN-CT-2000-00099 (BBW) and by the Russian Foundation
43: of Basic Research grant 02-01-22004.}}
44:
45: \date{\today}
46:
47:
48: \begin{document}
49:
50: \maketitle
51:
52:
53: \begin{abstract}
54: Enumerating ramified coverings of the sphere with fixed
55: ramification types is a well-known problem first considered
56: by A.~Hurwitz~\cite{Hurwitz}. Up to now, explicit solutions
57: have been obtained only for some families of ramified coverings,
58: for instant, those realized by polynomials in one complex
59: variable. In this paper we obtain an explicit answer for a large new
60: family of coverings, namely, the coverings realized by
61: simple almost polynomials, defined below. Unlike most other
62: results in the field, our formula is obtained by elementary methods.
63: \end{abstract}
64:
65:
66: \section{Rational functions and minimal factorizations of
67: permutations}
68:
69: Let $f: \C \rightarrow \C$ be a rational function
70: of degree $n$ in one complex variable.
71:
72: A {\em critical point} of $f$ is a point $z \in \C$ such that
73: $f'(z)=0$. Its {\em degree} is the number $a \geq 2$ such that $f$
74: looks like $f(z) = z^a$ in the neighborhood of the critical point.
75: A {\em critical value} of $f$ is its value at
76: a critical point. (Note that we do not count poles
77: as critical points.)
78:
79: \begin{definition}
80: A rational function $f$ is called {\em simple} if every
81: critical value of $f$ has exactly one critical preimage. It is
82: called {\em almost polynomial} if the sum of orders of its
83: poles is smaller than the degree of each
84: critical point.
85: \end{definition}
86:
87: Thus an almost polynomial rational function has a numerator
88: of a much bigger degree than the denominator.
89:
90: Our goal is to find the number of simple almost polynomial
91: rational functions with fixed orders of poles and
92: fixed critical values of fixed
93: degrees, a problem first considered by V.~Arnold~\cite{Arnold}.
94:
95: \bigskip
96:
97: The general problem of enumerating ramified coverings of the sphere
98: with fixed ramification types can, in some sense, be solved
99: using the representation theory of the symmetric group; however
100: the answer one obtains is a rather complicated sum over the
101: irreducible representations. In purticular, there is still no
102: simple criterion allowing one to determine whether the number of coverings
103: is equal to $0$ or not.
104:
105: For (not necessarily simple)
106: polynomials the problem can be reduced to a combinatorial problem
107: solved by I.~P.~Goulden, D.~M.~Jackson
108: in~\cite{GouJac}. Later, when the relation to polynomials was
109: discovered, their formula was reproved in
110: \cite{LanZvo}, \cite{PanZvo}, and~\cite{knizhka}
111: (appendix by D.~Zagier) by three different methods.
112: In~\cite{PanZvo} we also treated the case of almost polynomial functions
113: with just one simple pole. In all these cases, as well as in
114: the case considered in the present note, one obtains simple closed answers.
115:
116: A.~Goupil and G.~Schaeffer~\cite{GouSch} generalized
117: Goulden and Jackson's result on polynomials to meromorphic
118: functions with a unique pole on Riemann surfaces of any genus,
119: but their answer is not as explicit.
120: For other results on the enumeration of ramified coverings
121: and their relation to the intersection theory on moduli
122: spaces see~\cite{ELSV}, \cite{Zvonkine} and the references
123: therein.
124:
125: \bigskip
126:
127: All rational functions we consider in the sequel are simple and
128: almost polynomial.
129:
130: A rational function $f$ can be viewed as a ramified covering
131: of the Riemann sphere by the Riemann sphere. Going around
132: a critical value in the image we obtain a permutation of
133: the sheets in the preimage (the monodromy of the covering).
134: It follows from Riemann's existence theorem
135: that the problem of counting rational
136: functions can be reformulated in terms of permutations
137: (for details see~\cite{knizhka}, chapter~2).
138:
139: Let $\sigma \in S_n$ be a permutation of $n = \deg f$ elements.
140:
141: \begin{definition}
142: A product $\sigma_k \dots \sigma_1 = \sigma$ is called a
143: {\em minimal factorization} of $\sigma$ if (i)~the group
144: generated by $\sigma_1, \dots, \sigma_k$ acts transitively
145: on the set $\{ 1, \dots, n \}$ and (ii)~the total number
146: of cycles in the permutations
147: $\sigma_1, \dots, \sigma_k, \sigma$ equals $kn-n+2$.
148: \end{definition}
149:
150: Here $\sigma$ corresponds to the monodromy of $f$ at
151: $\infty$, while the $\sigma_j$'s are the monodromies
152: at the critical values. The conditions guarantee that the
153: ramified covering determined by the permutations
154: $\sigma_1, \dots, \sigma_k, \sigma$ is (i)~connected and
155: (ii)~of genus~$0$ (by the Riemann-Hurwitz formula).
156: Rather than counting the cycles of the permutations,
157: it is more natural to consider their {\em defects}:
158: a defect being $n-\mbox{(the number of cycles)}$.
159: If the Euler characteristic of the covering surface
160: equals $\chi$, the sum of defects of the corresponding
161: monodromies equals $2n - \chi$. Hence
162: the total defect $2n-2$ of a minimal factorization is
163: the smallest possible for a transitive factorization, which explains
164: the word ``minimal''.
165:
166: For shortness, we will say that {\em a
167: rational function $f$ is of type} $(a_1, \dots, a_k)$,
168: $(p_1, \dots, p_c)$ if it has $c$ poles of orders
169: $p_1, \dots, p_c$ and $k$ critical points of degrees
170: $a_1, \dots, a_k$. (The sum $p=p_1 + \dots + p_c$ satisfies
171: $p < a_j$ for all~$j$.) Similarly, we say that {\em a
172: minimal factorization is of type} $(a_1, \dots, a_k)$,
173: $(p_1, \dots, p_c)$ if $\sigma$ has cycles of lengths
174: $p_1, \dots, p_c, n-p$, while each $\sigma_j$ is an $a_j$-cycle,
175: i.e., has exactly one cycle of length $a_j$ the other points
176: being fixed.
177:
178: To a minimal factorization of type $(a_1, \dots, a_k)$,
179: $(p_1, \dots, p_c)$ we can assign a colored graph with oriented
180: edges called a {\em constellation}. It is obtained in the following
181: way. Take $n$ numbered vertices. For each $j$, $1 \leq j \leq k$
182: form an oriented polygon using the vertices from the cycle
183: of $\sigma_j$. The edges of the polygon are colored in ``color'' $j$.
184: Now forget the numbers of the vertices.
185:
186: It is clear that the
187: constellation allows one to reconstitute the permutations
188: $\sigma_1, \dots, \sigma_k, \sigma$ up to a common conjugation.
189:
190: \begin{definition} \label{Def:constellation}
191: A {\em constellation} is a connected graph whose edges are oriented and
192: colored in colors from $1$ to $k$, obtained by gluing together
193: $k$ oriented polygons with colors $1, \dots, k$ at some
194: of their vertices. A vertex of a polygon cannot be glued
195: to another vertex of the same polygon.
196: \end{definition}
197:
198: To sum up, we now have three equivalent problems: given
199: $(a_1, \dots, a_k)$, $(p_1, \dots, p_c)$, count
200: the number of rational functions,
201: or of minimal factorizations, or of constellations
202: of this type. The three numbers differ by simple combinatorial
203: factors.
204:
205: \bigskip
206:
207: We will need some more remarks on constellations.
208:
209: A constellation coming from a minimal
210: factorization has a natural embedding into a sphere
211: (see~\cite{knizhka}). It is
212: given by the following conditions: (i)~The orientations of the
213: edges determine the counterclockwise orientation on
214: each polygon; (ii)~If we choose any vertex and enumerate the
215: colors of the polygons surrounding it in the counterclockwise order
216: starting from the smallest color we obtain an increasing
217: sequence of colors.
218:
219: Cutting the sphere along the edges of the embedded constellation
220: we obtain $k+c+1$ pieces homeomorphic to open discs. Among
221: them, $k$ correspond to the polygons and the $c+1$ others
222: to the cycles of $\sigma$. The piece corresponding to the
223: long cycle (of length $n-p$) will be called the {\em
224: exterior face}; the other $c$ pieces will be called
225: {\em interior faces} or just {\em faces}. Note that
226: not all vertices of a face are contained in the corresponding
227: cycle of $\sigma$. Those which are will be called the
228: {\em essential vertices}, see Figure~\ref{Fig:faces}.
229:
230: \begin{figure}[h]
231: \begin{center}
232: \
233: \epsfbox{faces.eps}
234:
235: \caption{The color of each polygon is marked inside it.
236: The essential vertices of the central face are shown in black;
237: these vertices form a cycle of $\sigma_5 \dots \sigma_1$.}
238: \label{Fig:faces}
239:
240: \end{center}
241: \end{figure}
242:
243: From now on we will always assume that the constellations
244: are embedded into a sphere.
245:
246: Note that the somewhat bizarre condition $p < a_j$ has a
247: simple interpretation for constellations: it means that
248: every polygon must have at least one edge on the exterior face.
249: However the algebraic meaning of this condition remains
250: mysterious.
251:
252: \section{The main theorem}
253:
254: Let $p_1, \dots, p_c$ be $c \geq 0$ positive integers
255: (orders of poles), $\sum p_i = p$.
256: Fix an integer $n > p$ (degree of the almost polynomial).
257: Let $a_1, \dots a_k$ be $k \geq 2$ more
258: positive integers (multiplicities of the critical points)
259: satisfying $a_j >p$ for all $j$ and $\sum a_j = n+k+c-1$
260: (the Riemann-Hurwitz formula). Denote by $|\Aut \{ p_1, \dots, p_c \}|$
261: the number of permutations $s$ of $c$ elements such that
262: $p_i = p_{s(i)}$ for all $i$. For instance,
263: $|\Aut \{ 4,4,3,2,2,2,2,2,1,1 \}| = 2! \cdot 1! \cdot 5! \cdot 2 !\,$.
264:
265: Consider a permutation $\sigma \in S_n$ with cycle type
266: $(p_1, \dots, p_c, n-p)$.
267:
268: \begin{theorem} \label{Thm:main}
269: The number of minimal factorizations of $\sigma$ into
270: $a_j$-cycles equals
271: $$
272: \frac{(k+c-2)!}{(k-2)!} \; p_1^2 \dots p_c^2 \; (n-p)^{k-1}.
273: $$
274: The number of constellations as well as the number
275: of rational functions with fixed critical
276: values of type $(a_1, \dots, a_k)$,
277: $(p_1, \dots, p_c)$ equals
278: $$
279: \frac{1}{|\Aut \{ p_1, \dots, p_c \}|} \;
280: \frac{(k+c-2)!}{(k-2)!} \; p_1 \dots p_c \; (n-p)^{k-2}.
281: $$
282: \end{theorem}
283:
284: The three assertions of the theorem are equivalent. Indeed,
285: the number of constellations and the number
286: of rational functions of a given type coincide, which follows from
287: Riemann's existence theorem (see~\cite{knizhka} for more
288: details). On the other hand,
289: to obtain a minimal factorization of $\sigma$ from
290: a given constellation, we must number
291: the vertices of the constellation in such a way that the product
292: $\sigma_k \dots \sigma_1$ equals $\sigma$. It is easy to
293: see that there are
294: $$
295: p_1 \dots p_c (n-p) \cdot |\Aut \{ p_1, \dots p_c \}|
296: $$
297: ways to do that. (This number is also the number of permutations
298: that commute with $\sigma$.)
299:
300: In the next two sections we prove the theorem for constellations.
301:
302: \section{Assembling a constellation}
303:
304: We are going to prove the assertion of the theorem on the number
305: of constellations. We start by labeling the faces of
306: the constellations so as to make them distinguishable, which
307: kills the $|\Aut|$ factor.
308: We must show that the number of such constellations with
309: labeled faces equals
310: $$
311: \frac{(k+c-2)!}{(k-2)!} \; p_1 \dots p_c \; (n-p)^{k-2}.
312: $$
313:
314: We proceed by induction on the number $c$ of
315: faces in the constellations.
316:
317: For $c=0$, the constellations have
318: no (interior) faces, i.e., they are ``trees'' glued of
319: a given set of polygons. Such constellations (also called
320: ``cacti'') were enumerated in~\cite{GouJac},
321: \cite{LanZvo}, and~\cite{PanZvo}.
322: The answer one obtains is $n^{k-2}$.
323:
324: Suppose the formula is established for constellations with
325: $\leq c$ faces and let us add one more face.
326: A {\em polygon} of the constellation is a {\em neighbor} of
327: the $(c+1)$st face if it has at least one edge in
328: common with this face. A {\em face} of the constellation
329: is a {\em neighbor} of the $(c+1)$st face if its bounding
330: polygons are all neighbors of this face.
331: The $(c+1)$st face can have any number $2 \leq m \leq k$ of
332: neighboring polygons and any number $0 \leq d \leq c$ of
333: neighboring faces as in Figure~\ref{Fig:neighbors}.
334: (Note that a polygon that has only one vertex in common
335: with the face is not considered a neighbor.) The
336: condition $p < a_j$ implies that each neighboring face
337: is bounded by exactly two polygons.
338:
339:
340: \begin{figure}[h]
341: \begin{center}
342: \
343: \epsfbox{neighbors.eps}
344:
345: \caption{The $(c+1)$st face in this figure has $4$
346: neighboring polygons and $5$ neighboring faces.}
347: \label{Fig:neighbors}
348: \end{center}
349: \end{figure}
350:
351: To construct a constellation with $c+1$ faces we must
352: make the following choices.
353:
354: {\bf 1.}~Choose the numbers $m$ and $d$.
355:
356: {\bf 2.}~Choose $m$ polygons among $k$ and $d$ faces among $c$
357: to be the neighbors of the $(c+1)$st face. This gives
358: $$
359: {k \choose m} {c \choose d}
360: $$
361: choices. Denote by $D \subset \{1,\dots, c \}$ the set of the
362: $d$ neighboring faces.
363:
364: {\bf 3.}~Form the $(c+1)$st face using the $m$ chosen polygons. This
365: is done in the following way. Denote the essential
366: vertices of the face by $V_1, \dots, V_{p_{c+1}}$.
367: We must describe the colors and the order of the
368: edges that will form the intervals between $V_i$ and
369: $V_{i+1}$ for each $i$. We claim that such a disposition
370: of edges is uniquely determined once we have (arbitrarily)
371: assigned to each of the $m$ polygons the interval
372: $V_i V_{i+1}$ where its first edge will appear (as we go around the face in
373: the clockwise direction). Indeed, the disposition of
374: the edges can be obtained as follows. (a)~For each polygon
375: assigned to the interval $V_i V_{i+1}$ take one edge of
376: its color. (b)~Order these edges in the increasing order
377: of colors. (c)~In the case if either the biggest color used in
378: $V_{i-1} V_i$ is smaller than the smallest color in the
379: list for $V_i V_{i+1}$ or if there are no polygons assigned to
380: $V_i V_{i+1}$, add, at the beginning of the list of colors
381: for $V_i V_{i+1}$, the last color used in $V_{i-1}V_i$.
382: An example of this algorithm is shown in Figure~\ref{Fig:formface}.
383:
384: \begin{figure}[h]
385: \begin{center}
386: \
387: \epsfbox{formface.eps}
388:
389: \caption{How to obtain a face from a list of colors assigned
390: to each interval $V_i V_{i+1}$.}
391: \label{Fig:formface}
392: \end{center}
393: \end{figure}
394:
395: It is easy to see that the vertices $V_i$ are indeed the
396: essential vertices and that this is the unique way to achieve
397: this.
398:
399: Thus the number of ways to form the $(c+1)$st face using $m$ given
400: polygons is $p_{c+1}^m$.
401:
402: {\bf 4.}~Now we must choose the positions of the $d$ faces
403: that were chosen to be the neighbors of the $(c+1)$st face.
404: There are $m$ ``clefts'' between the polygons where these faces
405: can be placed. Several faces can appear in the same cleft.
406: In this case they must be ordered (starting from the face
407: closest to the $(c+1)$st one). Thus there are
408: $$
409: \frac{(d+m-1)!}{(m-1)!}
410: $$
411: ways to choose the positions of the faces.
412:
413: {\bf 5.}~Each of the neighboring faces has exactly two bounding
414: polygons. We now choose how many edges of the face will belong to
415: the polygon, say, on its left. For the $i$th face there are
416: $p_i$ choices. Thus we obtain a factor
417: $$
418: \prod_{i \in D} p_i .
419: $$
420: This product, of course, depends on the particular choice
421: of the $d$ neighboring faces. However we will soon see
422: that the remaining part of the formula contains the
423: complementary product
424: $$
425: \prod_{i \not\in D} p_i ,
426: $$
427: and thus it is not necessary to take a sum over all
428: possible choices.
429:
430: {\bf 6.}~We have assembled together all the
431: polygons and faces that are neighbors of the $(c+1)$st face. They
432: form a subconstellation $K$ of the total constellation. Now
433: we are going to {\em consider $K$ as a
434: unique polygon}. Indeed, we are going to show that the
435: remaining polygons are attached to $K$
436: in the same way as they would be attached to a unique polygon.
437: In particular, the number of ways to attach them is the same.
438: This will allow us to procede by induction.
439:
440: Denote by $r = p_{c+1}+\sum_{i \in D} p_i$ the sum of lengths
441: of the faces of $K$ and by $v$ the total number of
442: vertices of $K$. These $v$ vertices are acted upon by $m$ cyclic
443: permutation corresponding to the polygons. The product
444: of these $m$ permutations splits the vertices into $d+2$ cycles
445: of lengths $p_i$, $i \in D$, $p_{c+1}$, and $v-r$.
446: The vertices of the last cycle will be called the
447: {\em essential exterior vertices} of $K$.
448:
449: Now consider a constellation $R$ formed by the $k-m$
450: polygons not used in $K$ and, in addition, one
451: $(v-r)$-gon of color $0$. Suppose that the faces of $R$ have
452: lengths $p_i$, $i \not\in D$. We are going to replace
453: the $(v-r)$-gon in $R$ by the constellation $K$.
454:
455: Let
456: $p' = p_1 + \dots + p_c$ and $p = p_1 + \dots + p_{c+1}$.
457:
458: {\bf 7.}~First we establish a one-to-one correspondence between
459: the vertices of the $(v-r)$-gon and the essential exterior
460: vertices of $K$ preserving their cyclic order. There
461: are $v-r$ ways of doing that, which will account for
462: a factor $v-r$ in the final formula.
463:
464: In principle, the number $v-r$ is different
465: for different choices of the $m$ polygons and the $d$
466: faces. However we will soon see that this number
467: appears in the total sum only as a linear factor.
468: Therefore it will be a posteriori justified to replace
469: it by its average over the possible choices of $m$ polygons
470: and $d$ faces.
471:
472: The average value $\left< v-r \right>$ of $v-r$ equals
473: $$
474: \frac{m}{k} (n+k+c) -m - d -p_{c+1} - \frac{d}{c} p'
475: = \frac{mn+mc-k p_{c+1}}{k} -d \, \frac{p'+c}{c}.
476: $$
477: Indeed, the total number of edges in the $k$ polygons
478: equals $n+k+c$ by the Euler formula. Choosing randomly $m$ of
479: the $k$ polygons we obtain an average of $\frac{m}{k}(n+k+c)$
480: edges in the subconstellation $K$. Since $K$
481: has $d+1$ faces and $m$ polygons, its average number
482: of vertices is $\left< v \right> = \frac{m}{k}(n+k+c) - m-d$,
483: again by the Euler formula.
484: Now, $K$ has $d+1$ faces, one of which is always of
485: length $p_{c+1}$ and the other $d$ are chosen randomly
486: from $c$ possibilities. Thus $ \left< r \right>
487: = p_{c+1} + \frac{d}{c} p'$, whence we obtain $\left< v-r \right>$.
488:
489: {\bf 8.}~Let us go around the exterior face
490: of the constellation $K$ in the counterclockwise direction.
491: The colors of the edges we meet will be increasing
492: in each interval between two consecutive essential exterior
493: vertices. Then, as we pass an essential vertex, the number of the
494: color jumps down.
495:
496: Suppose we are given a new polygon of some color $j$ that
497: does not appear in the constellation $K$. We want to
498: attach this polygon to the exterior of $K$ in such a way
499: that the cyclic order of colors around each vertex remains
500: increasing (see the remarks after
501: Definition~\ref{Def:constellation}). It is easy to
502: see that in each interval between two consecutive
503: essential vertices $V_i$ (included) and $V_{i+1}$ (excluded)
504: there is a unique vertex to which the polygon can be attached.
505:
506: Now, the essential exterior vertices of $K$ are in
507: a one-to-one correspondence with the vertices of a
508: $(v-r)$-gon in the new constellation $R$. We want to
509: replace the $(v-r)$-gon in $R$ by the subconstellation $K$.
510: To do that, for each polygon $j$ attached to some vertex
511: of the $(v-r)$-gon, we take the corresponding essential
512: exterior vertex $V_i$ of $K$ and attach our polygon to
513: the unique possible vertex between $V_i$ and $V_{i+1}$.
514:
515: It is easy to see that the faces of $R$, even those that
516: have been modified by our operation, will still have the
517: same number of essential vertices as before. Thus there
518: is a unique way to substitute the $(v-r)$-gon in $R$
519: by the subconstellation $K$. The result is the constellation
520: we were trying to assemble.
521:
522: {\bf 9.}~It remains to choose the constellation $R$.
523: By the induction assumption, there are
524: $$
525: \frac{(k-m-1+c-d)!}{(k-m-1)!} (n-p)^{k-m-1}
526: \prod_{i \not\in D} p_i
527: $$
528: choices. Indeed, the constellation $R$ has $k-m$ polygons,
529: $c-d$ (interior) faces of lengths $p_i$, $i \not\in D$,
530: and the length of its exterior face is
531: $n-p$ (the same as in the constellation that we
532: are assembling).
533:
534:
535: \section{Computing the sum}
536:
537: The result of our investigation is that
538: the number of constellations (with labeled faces)
539: is given by the following sum:
540:
541: $$
542: S=
543: p_1 \dots p_c
544: \sum_{m=2}^k {k \choose m} (n-p)^{k-m-1} p_{c+1}^{m-1}
545: \times \qquad \qquad \qquad \qquad \qquad \qquad
546: $$
547: $$
548: \times \sum_{d=0}^c {c \choose d}
549: \frac{(d+m-1)!}{(m-1)!}
550: \frac{(k-m-1+c-d)!}{(k-m-1)!}
551: \left(
552: \frac{mn \! + \! mc \! - \! k p_{c+1}}{k} -d \, \frac{p'+c}{c}
553: \right).
554: $$
555:
556: The rest of the proof is a sequence of elementary but rather
557: cumbersome computations.
558:
559: We first compute the subsum over $d$ using the elementary relations
560: $$
561: \sum_{d=0}^c {a+d \choose d}{b-d \choose c-d} = {a+b+1 \choose c},
562: $$
563: $$
564: \sum_{d=0}^c d {a+d \choose d}{b-d \choose c-d} =
565: (a+1) {a+b+1 \choose c-1}
566: $$
567: with $a = m-1$, $b = k+c-m-1$.
568: We find that the subsum is equal to
569: $$
570: \frac{(k+c-1)!}{(k-1)!}
571: \left( \frac{mn+mc-k p_{c+1}}{k} \right)
572: -m \frac{(k+c-1)!}{k!} (p'+c)
573: $$
574: $$
575: = \frac{(k+c-1)!}{k!} [ m(n-p') - k p_{c+1}].
576: $$
577:
578: Substituting this into the initial sum we obtain
579: $$
580: S = p_1 \dots p_c \frac{(k+c-1)!}{k!}
581: \sum_{m=2}^k {k \choose m} (n-p)^{k-m-1} p_{c+1}^{m-1}
582: [m(n-p') - k p_{c+1}].
583: $$
584: This sum can now be evaluated using two more elementary
585: identities
586: $$
587: \sum_{m=2}^k m {k \choose m} x^{k-m-1} y^{m-1}
588: = k \left[ \frac{(x+y)^{k-1}}{x} - x^{k-2} \right]
589: $$
590: $$
591: \sum_{m=2}^k {k \choose m} x^{k-m-1} y^m
592: = \frac{(x+y)^k}{x} - x^{k-1} - kx^{k-2}y,
593: $$
594: with $x = n-p$, $y = p_{c+1}$. We obtain
595: $$
596: S = p_1 \dots p_c \frac{(k+c-1)!}{k!}
597: \left\{
598: k \left[
599: \frac{(n-p')^{k-1}}{n-p} - (n-p)^{k-2}
600: \right] (n-p') -
601: \right.
602: \qquad \qquad
603: $$
604: $$
605: \qquad \qquad
606: \left.
607: -k \left[
608: \frac{(n-p')^k}{n-p} - (n-p)^{k-1} - k p_{c+1} (n-p)^{k-2}
609: \right]
610: \right\}
611: $$
612: $$
613: = p_1 \dots p_c \frac{(k+c-1)!}{(k-1)!}
614: (n-p)^{k-2}
615: \left\{
616: -(n-p')+(n-p)+kp_{c+1}
617: \right\}
618: $$
619: $$
620: = p_1 \dots p_{c+1} \frac{(k+c-1)!}{(k-2)!} (n-p)^{k-2}.
621: $$
622:
623: This is precisely the formula of Theorem~\ref{Thm:main}
624: with $c$ replaced by $c+1$. The theorem is proved.
625:
626: \begin{thebibliography}{99}
627:
628: \bibitem{Arnold} {\bf V. I. Arnold.}
629: {\em Topological classification of complex trigonometric polynomials and
630: the combinatorics of graphs with an identical number of vertices and
631: edges.} (Russian) Funktsionalnyi Analiz i Prilozheniia, {\bf 30} (1996),
632: No.~1, 1--17, 96; translation in Functional Analysis and Applications,
633: {\bf 30} (1996), No.~1, 1--14.
634:
635: \bibitem{ELSV} {\bf T. Ekedahl, S. K. Lando, M. Shapiro,
636: A. Vainshtein.} {\em Hurwitz numbers and intersections
637: on moduli spaces of curves.} -- Inventiones Mathematicae,
638: {\bf 146} (2001), 297--327, {\tt arxiv:math.AG/0004096}.
639:
640: \bibitem{GouJac} {\bf I. P. Goulden, D. M. Jackson.}
641: {\em The combinatorial relationship between trees, cacti
642: and certain connection coefficients for the symmetric group.}
643: -- European Journal of Combinatorics, {\bf 13} (1992), 357--365.
644:
645: \bibitem{GouSch} {\bf A. Goupil, G. Schaeffer.}
646: {\em Factoring $n$-cycles and counting maps of given
647: genus.}
648: -- European Journal of Combinatorics, {\bf 19} (1998), No.~7,
649: 819--834.
650:
651: \bibitem{Hurwitz} {\bf A. Hurwitz.}
652: {\em \"Uber Riemann'sche Fl\"achen mit gegebenen
653: Verzweigungs\-punkten.} -- Mathematische Annalen, {\bf 39} (1891),
654: 1--61.
655:
656: \bibitem{knizhka} {\bf S. K. Lando, A. K. Zvonkin.}
657: {\em Graphs on Surfaces and Their Applications.}
658: -- Springer-Verlag, 2004.
659:
660: \bibitem{LanZvo} {\bf S. K. Lando, D. Zvonkine.}
661: {\em On multiplicities of the Lyashko-Looijenga mapping
662: on the discriminant strata.}
663: -- Functional Analysis and its Applications,
664: {\bf 33} (2000), 178--188.
665:
666: \bibitem{PanZvo} {\bf D. Panov, D. Zvonkine.}
667: {\it Counting meromorphic functions with critical points
668: of large multiplicities.}
669: -- Notes of the Scientific Seminar of the Mathematical
670: Department of St-Petersburg Steklov Institute (POMI),
671: {\bf 292} (2002), Representation Theory, Dynamical Systems,
672: Combinatorics, and Algorithmic Methods, No.~7, 92--119.
673: Available on {\tt arXiv: math.CO/0209013}.
674:
675: \bibitem{Zvonkine} {\bf D. Zvonkine.}
676: {\em An algebra of power series arising in the intersection
677: theory of moduli spaces of curves and in the enumeration of
678: ramified coverings of the sphere.}
679: -- {\tt arXiv:math.AG/0403092} (2004).
680:
681: \end{thebibliography}
682:
683: \end{document}
684: