1: %%%%%%%%%%%%%%%%%%%%%%%% Ams-Style %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
2: %%%
3: %%% Style and Inputs
4: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
5:
6: \documentclass[10pt]{amsart}
7: \usepackage{amssymb}
8: \usepackage{clrscode}
9: \usepackage{epsfig}
10: \usepackage{latexsym}
11: \usepackage{amsmath,amssymb,amsfonts,amsthm,graphics}%,latexcad
12: \usepackage{eucal}% caligraphic-euler fonts: \mathcal{ }
13: \usepackage{eufrak}% frak-euler fonts: \mathfrak{ }
14: \usepackage[all]{xypic}
15: \usepackage{xspace}
16: %\usepackage{layout}% displays settings; use \layout in text
17:
18: %%%
19: %%%
20: %%%%%%%%%%%%%%%%%%%%%%%%% Pagestyle %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
21: %%%
22: %%%
23:
24: \renewcommand{\baselinestretch}{1.2}% spacing between lines
25:
26: %\hoffset=0truecm
27: %\voffset=0truecm
28: \textwidth=15truecm \textheight=18truecm \baselineskip=0.8truecm
29: \overfullrule=0pt
30: \parskip=0.8\baselineskip
31: \parindent=0truecm
32: \topmargin=0.5truecm \headsep=1.2truecm
33: %\oddsidemargin=0.5in % options for double-side printouts
34: %\evensidemargin=0in
35:
36: %%%
37: %%%
38: %%%%%%%%%%%%%%%%%%%% New Settings %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
39: %%%
40: %%%
41:
42: \theoremstyle{plain}
43: \newtheorem{theorem}{Theorem}
44: \newtheorem{corollary}{Corollary}
45: \newtheorem*{main}{Main~Theorem}
46: \newtheorem{lemma}{Lemma}
47: \newtheorem{proposition}{Proposition}
48:
49: \theoremstyle{definition}
50: \newtheorem{definition}{Definition}
51:
52: \theoremstyle{remark}
53: \newtheorem{remark}{Remark}
54: \newtheorem*{notation}{Notation}
55: \newtheorem{example}{Example}
56: \numberwithin{equation}{section}
57:
58:
59: \begin{document}
60: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
61:
62: %%%
63: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
64: %%%%%
65:
66:
67: \title[Asymptotic analysis of $k$-noncrossing matchings]
68: {Asymptotic analysis of $k$-noncrossing matchings}
69: \author{Emma Y. Jin$^{}$, Christian M. Reidys$^{}$$^{\,\star}$
70: and Rita R. Wang $^{}$}
71: \address{ Center for Combinatorics, LPMC-TJKLC \\
72: Nankai University \\
73: Tianjin 300071\\
74: P.R.~China\\
75: Phone: *86-22-2350-6800\\
76: Fax: *86-22-2350-9272}
77: \email{reidys@nankai.edu.cn}
78: \thanks{}
79: \keywords{determinant, Bessel-function, subtraction of singularity principle,
80: oscillating tableaux}
81: \date{March, 2008}
82: \begin{abstract}
83: In this paper we study $k$-noncrossing matchings. A $k$-noncrossing
84: matching is a labeled graph with vertex set $\{1,\dots,2n\}$
85: arranged in increasing order in a horizontal line and vertex-degree
86: $1$. The $n$ arcs are drawn in the upper halfplane subject to the
87: condition that there exist no $k$ arcs that mutually intersect. We
88: derive: (a) for arbitrary $k$, an asymptotic approximation of the
89: exponential generating function of $k$-noncrossing matchings
90: $F_k(z)$. (b) the asymptotic formula for the number of
91: $k$-noncrossing matchings $f_{k}(n) \, \sim \,c_k \,
92: n^{-((k-1)^2+(k-1)/2)}\, (2(k-1))^{2n}$ for some $c_k>0$.
93: \end{abstract}
94: \maketitle {{\small
95: %\tableofcontents
96: }}
97:
98:
99: %%%
100: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
101: %%%
102: %%%
103:
104: \section{Statement of results and background}\label{S:intro}
105:
106: %%%
107: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
108: %%%
109:
110: Let $F_k(z)$ denote the exponential generating function of $k$-noncrossing
111: matchings, i.e.~
112: \begin{equation}
113: F_k(z)=\sum_{n\ge 0}f_k(n)\, \frac{z^{2n}}{(2n)!} \ .
114: \end{equation}
115: In this paper we prove the following two theorems:
116: %%%
117: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
118: %%%
119: \begin{theorem}\label{T:A}
120:
121: Then we have for arbitrary $k\in\mathbb{N}$, $k\ge 2$, $\text{\rm
122: arg($z$)} \ne \pm\frac{\pi}{2}$ %and $\vert z\vert\rightarrow \infty$
123: \begin{eqnarray}\label{E:das1}
124: F_k(z) & = &
125: \left[\prod_{i=1}^{k-1}\Gamma(i+1-\frac{1}{2}) \, \prod_{r=1}^{k-2}r!\right]
126: \left(\frac{e^{2z}}{\pi}\right)^{k-1}\, z^{-(k-1)^2-\frac{k-1}{2}}\;
127: \ (1+{O}(|z|^{-1})) \ .
128: \end{eqnarray}
129: \end{theorem}
130: %%%
131: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
132: %%%
133: %%%
134: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
135: %%%
136: \begin{theorem}\label{T:B}
137: For arbitrary $k\in\mathbb{N}$, $k\ge 2$ we have
138: \begin{equation}\label{E:theorem}
139: f_{k}(n) \, \sim \, c_k \, n^{-((k-1)^2+(k-1)/2)}\,
140: (2(k-1))^{2n},\qquad \text{\it for some $c_k>0$} \ .
141: \end{equation}
142: \end{theorem}
143: %%%
144: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
145: %%%
146: Here we use the notation $f(z)=O(g(z))$ and $f(z)=o(g(z))$ for
147: $|f(z)|/|g(z)|$ being bounded and tending to zero, for $\vert z
148: \vert\rightarrow \infty$, respectively.
149:
150: A $k$-noncrossing matching is a labeled graph over the vertices $1,\dots,2n$,
151: of degree exactly $1$ and drawn in increasing order in a horizontal line.
152: The arcs are drawn in the
153: upper halfplane subject to the condition that there are no $k$ arcs that
154: mutually intersect.
155: %%%
156: %%%%%%%%%%%%%%%%%%%%%%%% Figures 2 %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
157: %%%
158: \begin{figure}[ht]
159: \centerline{ \epsfig{file=match-k.eps,width=1.0\textwidth}\hskip15pt
160: } \caption{\small $k$-noncrossing matchings: $3$-, $4$-, $5$- and
161: $6$-noncrossing matchings respectively. One of the $k-1$ mutually
162: crossing arcs are drawn in red.} \label{F:diagram}
163: \end{figure}
164: %%%
165: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
166: %%%
167: Grabiner and Magyar proved an explicit determinant formula,
168: \cite{GM} (see also \cite{CDDSY}, eq.~$(9)$) which expresses the
169: exponential generating function of $f_k(n)$, for fixed $k$, as a
170: $(k-1) \times (k-1)$ determinant
171: \begin{eqnarray}
172: \label{E:ww1} F_k(z)=\sum_{n\ge 0} f_{k}(n)\cdot\frac{z^{2n}}{(2n)!}
173: & = & \det[I_{i-j}(2z)-I_{i+j}(2z)]|_{i,j=1}^{k-1} \ ,
174: \end{eqnarray}
175: where $I_m(2z)$ is the hyperbolic Bessel function:
176: \begin{equation}\label{E:bessel}
177: I_m(2z)=\sum_{j=0}^{\infty} { {z^{m+2j}} \over {j!(m+j)!} } \quad .
178: \end{equation}
179: Chen {\it et.al.} proved in \cite{CDDSY} a beautiful correspondence
180: between $k$-noncrossing matchings and oscillating tableaux. The
181: particular RSK-insertion used in \cite{CDDSY} is based on an idea of
182: Stanley.
183: Our second result is related to a theorem of Regev \cite{Regev:81}
184: for the coefficient $u_k(n)$ of Gessel's generating function \cite{Gessel}
185: $$
186: U_k(x)=\det(I_{i-j}(2z))_{i,j=1}^k \ .
187: $$
188: Regev shows
189: \begin{equation}
190: u_k(n)
191: \sim 1!2!\dots (k-1)!\left(\frac{1}{\sqrt{2\pi}}\right)^{k-1}\,
192: \left(\frac{1}{2}\right)^{(k-1)^2/2} k^{k^2/2}\frac{k^{2n}}{n^{(k^2-1)/2}} \ .
193: \end{equation}
194: The proof is obtained employing the RSK-algorithm and using the
195: hook-length formula. One arrives, taking the limit $n\to\infty$, at
196: a $k$-dimensional {\it Selberg}-integral, which can be explicitely
197: computed. We shall use a different strategy. One key element in our
198: approach is the following approximation of the Bessel-function,
199: valid for $-\frac{\pi}{2}<
200: \text{\rm arg(z)} < \frac{\pi}{2}$ %and $\vert z \vert\rightarrow\infty$
201: \cite{NBS:70}
202: \begin{equation}\label{E:bessel2}
203: I_m(z)=\frac{e^{z}}{\sqrt{2 \pi
204: z}}\left(\sum_{h=0}^{H}\frac{(-1)^{h}}{h!8^h}\prod\limits_{t=1}^h
205: (4m^2-(2t-1)^2)z^{-h}+\mathrm{O}(|z|^{-H-1})\right) \ .
206: \end{equation}
207: In this paper we will show that, using the approximation of
208: eq.~(\ref{E:bessel2}), the determinant of Bessel-functions of
209: eq.~(\ref{E:ww1}) can be computed asymptotically for arbitrary $k$.
210: The computation of the determinant via the algorithm given in
211: Section~\ref{S:Proof-A} is the key ingredient for all our results.
212:
213:
214: %%%
215: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
216: %%%
217:
218: \section{Proof of Theorem~\ref{T:A}}\label{S:Proof-A}
219:
220: %%%
221: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
222: %%%
223: Suppose we are given a polynomial
224: \begin{equation}\label{E:def-f}
225: g_n(x,y) = \sum_{0\leq a+b\leq n}C(a,b)\, x^{2a}y^{2b} \ ,
226: \end{equation}
227: where for $a+b=n$, $C(a,b)>0$ holds.
228: In the following $a$, $b$ always denote integers greater or equal to zero.
229: We set
230: \begin{equation}\label{E:triangle}
231: z\triangle z'= (z-z')(z+z') \ .
232: \end{equation}
233: %%%
234: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
235: %%%
236: \begin{lemma}\label{L:polynomial}
237: Suppose $n\ge 0$, then we have
238: \begin{equation}\label{E:(a)}
239: g_n(x,y)-g_n(x,z) =
240: \begin{cases}
241: (y\triangle z)\,
242: \sum\limits_{a+b\leq n-1}E(a,b,z)\, x^{2a}\,y^{2b} & n\ge 1 \\
243: 0 & n=0
244: \end{cases}
245: \end{equation}
246: where $E(a,b,z)=C(a,b+1)$ for $a+b=n-1$. Furthermore
247: \begin{equation}\label{E:(b)}
248: g_n(x,y)-g_n(x,y_1)-g_n(x_1,y)+g_n(x_1,y_1)=
249: \begin{cases}
250: (x\triangle x_1)(y\triangle y_1)\sum\limits_{ a+b\leq
251: n-2}D(a,b,x_1,y_1)x^{2a}y^{2b} & n\geq 2 \\
252: 0 & n=0,1
253: \end{cases}
254: \end{equation}
255: where $D(a,b,x_1,y_1)=C(a+1,b+1)$ for $a+b=n-2$.
256: \end{lemma}
257: %%%
258: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
259: %%%
260: \begin{proof}
261: For $n=0$, we immediately obtain $g_0(x,y)-g_0(x,z)=0$. In case of
262: $n\geq 1$ we compute
263: \begin{align*}
264: g_n(x,y)-g_n(x,z) &= (y\triangle z)\sum\limits_{ a+b\leq n, \,b>0}
265: C(a,b)\, x^{2a}\left[\sum_{m=0}^{b-1}\, y^{2m}z^{2b-2-2m}\right]\\
266: &=(y\triangle z)\sum\limits_{ a+b\leq n-1}E(a,b,z)\, x^{2a}y^{2b}.
267: \end{align*}
268: In particular, for $a+b=n-1$, we observe $E(a,b,z)=C(a,b+1)$. As for
269: eq.~(\ref{E:(b)}) we compute in case of $n=0\ \text{or}\ 1$,
270: $\vartheta_n(x,x_1,y,y_1)=g_n(x,y)-g_n(x,y_1)-g_n(x_1,y)+g_n(x_1,y_1)=0$.
271: For $n\geq 2$ we compute
272: \begin{eqnarray*}
273: \vartheta_n(x,x_1,y,y_1)&=& (x\triangle x_1)(y\triangle
274: y_1)\sum\limits_{a+b\leq n, \, ab>0}
275: C_h(a,b)\left[\sum_{m=0}^{a-1}x^{2m}x_1^{2a-2-2m}\right]
276: \left[\sum_{m=0}^{b-1}y^{2m}y_1^{2b-2-2m}\right]\\
277: &=&(x\triangle x_1)(y\triangle y_1)\sum_{a+b\leq n-2}
278: D(a,b,x_1,y_1)\, x^{2a}y^{2b} \ .
279: \end{eqnarray*}
280: In particular, for $a+b=n-2$, $D(a,b,x_1,y_1)=C(a+1,b+1)$ holds.
281: \end{proof}
282: %%%
283: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
284: %%%
285: Let
286: \begin{eqnarray}\label{E:e}
287: e_{i,j}(z) & = & \sum_{h\geq 0}m_h(i,j)\frac{(-1)^h}{16^hh!}\, z^{-h}\\
288: \label{E:m}
289: m_h(i,j) & = & \prod_{t=1}^{h}(4(i-j)^2-(2t-1)^2)-
290: \prod_{t=1}^{h}(4(i+j)^2-(2t-1)^2) \ .
291: \end{eqnarray}
292: We consider the algorithm {\bf A}, specified in Figure~\ref{F:algo}, which
293: manipulates the matrix of Laurent series
294: $M=(e_{i,j}(z))_{1\le i,j\le k-1}$.
295: %%%
296: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
297: %%%
298: \begin{figure}
299: %{\small
300: \begin{codebox}
301: \zi $\qquad$ $\qquad$$\qquad$ $\qquad$$\qquad$ $\qquad$$\qquad$ $\qquad$
302: $\qquad$ $\qquad$\centerline{\bf The algorithm {\bf A}: }
303: \zi \kw{begin}
304: \zi $M:=[e_{i,j}(z)]_{i,j=1}^{k-1};$
305: \zi $\quad$ $\quad$ \For $t$ from $1$ to $k-1\ \kw{do}$
306: \zi $\quad$$\qquad$ $\qquad$ \For $i$ from $t+1$ to $k-1$ \kw{do}
307: \zi $\quad$$\quad$$\qquad$ $\quad$$\qquad$ \For $j$ from $1$ to $k-1$ \kw{do}
308: \zi $\quad$$\quad$$\qquad$ $\quad$$\qquad$ $e_{i,j}'(z):=
309: \frac{-i\,
310: \prod_{r=1}^{t-1}(i\triangle r)}{(2t-1)!}e_{t,j}(z)+e_{i,j}(z);$
311: \zi $\quad$$\quad$$\qquad$ $\quad$$\qquad$ $e_{i,j}(z):=e_{i,j}'(z);$
312: \zi $\quad$$\quad$$\qquad$ $\quad$$\qquad$ \kw{end};
313: \zi $\quad$$\qquad$ $\qquad$ \kw{end};
314: \zi $\quad$$\qquad$ $\qquad$ \For $j$ from $t+1$ to $k-1$ \kw{do}
315: \zi $\quad$$\quad$$\qquad$ $\quad$$\qquad$ \For $i$ from $1$ to $k-1$ \kw{do}
316: \zi $\quad$$\quad$$\qquad$ $\quad$$\qquad$ $e_{i,j}''(z):=\frac{-j\,
317: \prod_{r=1}^{t-1}(j\triangle r)}{(2t-1)!}e_{i,t}(z)+e_{i,j}(z);$
318: \zi $\quad$$\quad$$\qquad$ $\quad$$\qquad$ $e_{i,j}(z):=e_{i,j}''(z);$
319: \zi $\quad$$\quad$$\qquad$ $\quad$$\qquad$ \kw{end};
320: \zi $\quad$$\qquad$ $\qquad$ \kw{end};
321: \zi $\quad$ $\quad$ \kw{end};
322: \zi \kw{output} $M$;
323: \zi \kw{end};
324: \end{codebox}
325: %}
326: \caption{\small }
327: \label{F:algo}
328: \end{figure}
329: %%%
330: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
331: %%%
332: Let $e_{i,j}^t(z)$ denote the matrix coefficient after running {\bf A} exactly
333: $t$ steps. We set
334: \begin{equation}
335: e_{i,j}^t(z) = \sum_{h\geq 0} m_h^t(i,j)\frac{(-1)^h}{16^hh!}\, z^{-h}
336: \end{equation}
337: and proceed by analyzing the terms $m_h^t(i,j)$ for $1\le t<k-1$.
338: %%%
339: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
340: %%%
341: \begin{lemma}\label{L:induction}
342: For any positive integer $t$ strictly smaller than $k-1$ and we have
343: $m_h^t(i,j)=m_h^t(j,i)$ the following
344: two assertions hold.\\
345: {\rm (a)} for $i\leq t < j$, we have
346: \begin{eqnarray}
347: m_h^t(i,j) & = & -(2i-1)!j\,\prod_{r=1}^t(j\triangle r)\,
348: \sum\limits_{ a+b\leq h-(t+i)} E_h^{t}(a,b,i) \, i^{2a}j^{2b} \\
349: h<t+i & \Longrightarrow & m_h^t(i,j)=0 \\
350: a+b=h-(t+i) & \Longrightarrow & E_h^t(a,b,i)=C_h(a+i-1,b+t)>0 \ .
351: \end{eqnarray}
352: Furthermore {\rm (a)} implies the case $j \le t<i$.
353:
354: {\rm (b)} for $i,j>t$, we have
355: \begin{eqnarray}
356: m_h^t(i,j) & = & -i\,j\; \prod_{r=1}^t (i\triangle r)\,
357: \prod_{r=1}^t (j\triangle r)\,
358: \sum\limits_{ a+b\leq h-(2t+1)} D_h^{t}(a,b)\, i^{2a}j^{2b}\\
359: h<2t+1 & \Longrightarrow & m_h^t(i,j)=0 \\
360: a+b=h-(2t+1) & \Longrightarrow & D_h^t(a,b)=C_h(a+t,b+t)>0 \ .
361: \end{eqnarray}
362: \end{lemma}
363: %%%
364: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
365: %%%
366: \begin{proof}
367: We shall prove (a) and (b) by induction on $1\le t<k-1$.
368: We first observe that, in view of eq.~(\ref{E:m})
369: \begin{equation}\label{E:qw}
370: \begin{split}
371: m_h(i,j) &=
372: -2\sum_{s=0}^{\lfloor\frac{h-1}{2}\rfloor}\sum_{p+q+r+2s+1=h}
373: \binom{h-r}{p}\binom{h-p-r}{q}(4i^2)^p(4j^2)^qa_1\cdots a_r(8ij)^{2s+1}\\
374: &= -ij\sum_{a+b\leq h-1}C_h(a,b)i^{2a}j^{2b} \ ,
375: \end{split}
376: \end{equation}
377: where $a_i\in\{-1^2,\ldots,-(2h-1)^2\}$, $i\neq j,\ a_i\neq a_j$ and
378: $C_h(a,b)>0$ for $a+b=h- 1$. Furthermore by definition
379: $m_h(i,j)=m_h(j,i)$. For $i=1,j>1$, only the $j$-loop is executed,
380: whence
381: \begin{equation}
382: m_h^1(i,j)= m_h(i,j)-jm_h(i,1)
383: \end{equation}
384: and for $m_h(j,i)$ only the $i$-loop contributes
385: $$
386: m_h^{1}(j,i)= m_h(j,i)-jm_h(1,i)= m_h(i,j)-jm_h(i,1)=m_h^{1}(i,j) \
387: .
388: $$
389: Consequently
390: $$
391: m_h^1(i,j)=-ij\left[ \sum_{a+b\leq
392: h-1}C_h(a,b)i^{2a}j^{2b}-\sum_{ a+b\leq h-1
393: }C_h(a,b)i^{2a}\right]
394: $$
395: Employing Lemma \ref{L:polynomial} we obtain
396: $$
397: m_h^1(i,j)=-i\, j\, (j\triangle 1)
398: \sum_{a+b\leq h-2}E_h^1(a,b,1)i^{2a}j^{2b} \ .
399: $$
400: Furthermore
401: \begin{eqnarray*}
402: a+b=h-2 & \Longrightarrow & E_h^1(a,b,1)=C_h(a,b+1)>0 \\
403: h= 1 & \Longrightarrow & m_h^1(i,j)=0 \ .
404: \end{eqnarray*}
405: Thus for $t=1$, the induction basis for (a) holds.
406: We proceed by proving that for $t=1$ (b) holds.
407: For $i>1,\ j>1$ both $i$- and $j$-loop are executed
408: \begin{eqnarray}
409: m_h^1(i,j)&=&m_h(i,j)-im_h(1,j)-jm_h(i,1)+ijm_h(1,1)
410: \end{eqnarray}
411: from which immediately $m_h^1(i,j)=m_h^1(j,i)$ follows. We compute
412: \begin{eqnarray*}
413: m_h^1(i,j) &=&-ij\left(\sum_{ a+b\leq h-1
414: }C_h(a,b)i^{2a}j^{2b}-\sum_{ a+b\leq h-1}C_h(a,b)j^{2b}\right)\\
415: &&-ij\left(-\sum_{ a+b\leq h-1 }C_h(a,b)i^{2a}+\sum_{
416: a+b\leq h-1}C_h(a,b)\right)\\
417: &=&-i(i\triangle 1)j(j\triangle 1)
418: \sum_{a+b\leq h-3 }D_h^{1}(a,b)i^{2a}j^{2b} \ .\\
419: \end{eqnarray*}
420: Lemma~\ref{L:polynomial} implies for $a+b=h-3$, $D_h^1(a,b)=C_h(a+1,b+1)$
421: and for $h<3$, $m_h^1(i,j)=0$.
422: Accordingly we eststablished the induction basis for assertions (a),
423: (b) and $m_h^1(i,j)=m_h^1(j,i)$.\\
424: As for the induction step, we first prove {\rm (a)}. Let us suppose
425: {\rm (a)} holds for $t=n$. We consider the case $t=n+1$ by
426: distinguishing subsequent two cases: {\rm (1)} $i=n+1, j>n+1$ and
427: {\rm (2)} $i\le n, j=n+1$. First we observe that since $i<n+2$ the
428: algorithm executes no $i$-loop and by construction the only
429: contribution to $m_h^{n+1}(i,j)$ is made by the term
430: $$
431: -\frac{j\prod_{r=1}^{n+1-1}(j\triangle r)}{(2n+1)!} \, m_h^n(i,n+1)
432: $$
433: We accordingly derive
434: \begin{eqnarray*}
435: m_h^{n+1}(i,j) &=&
436: m_h^n(i,j)-\frac{j\prod_{r=1}^n(j\triangle r)}{(2n+1)!}m_h^n(i,n+1)
437: \end{eqnarray*}
438: The induction hypothesis on $t=n$ shows $m_h^n(i,j)=m_h^n(j,i)$
439: and $m_h^n(i,n+1)=m_h^n(n+1,i)$. Therefore we arrive at
440: $m_h^{n+1}(i,j)=m_h^{n+1}(j,i)$.
441:
442: {\rm (1)} $i=n+1, j>n+1$, the induction hypothesis guarantees
443: \begin{eqnarray*}
444: m_h^{n}(i,j) & = & -i\,j\; \prod_{r=1}^n (i\triangle r)\,
445: \prod_{r=1}^{n} (j\triangle r)\,
446: \sum\limits_{ a+b\leq h-(2n+1)} D_h^{n}(a,b)\, i^{2a}j^{2b}\\
447: m_h^n(i,n+1) & = & -i\,(n+1)\; \prod_{r=1}^n (i\triangle r)\,
448: \prod_{r=1}^{n} ((n+1)\triangle r)\, \sum\limits_{a+b\leq
449: h-(2n+1)} D_h^{n}(a,b)\, i^{2a}(n+1)^{2b}
450: \end{eqnarray*}
451: Since $(n+1)\, \prod_{r=1}^{n}((n+1)\triangle r)=(2n+1)!$
452: we arrive at
453: \begin{eqnarray*}
454: m_h^{n+1}(i,j)
455: &=&- i\,j\, \prod_{r=1}^n(i\triangle r)\, \prod_{r=1}^n(j\triangle r)
456: \times \\
457: &&
458: \left(\sum_{a+b\leq h-(2n+1)}D_h^n(a,b)i^{2a}j^{2b}-
459: \sum_{a+b\leq h-(2n+1)} D_h^n(a,b)i^{2a}(n+1)^{2b}\right)\ .\\
460: \end{eqnarray*}
461: According to Lemma \ref{L:polynomial}, for $h=2n+1$, we have
462: $m_h^{n+1}(n+1,j)=0$ and for $h\ge 2n+2$ we obtain
463: \begin{eqnarray}
464: \quad m_h^{n+1}(n+1,j) &=&-(2n+1)!j\prod_{r=1}^{n+1}(j\triangle r )
465: \sum_{a+b\leq h-(2n+2)}E_h^{n+1}(a,b,n+1)(n+1)^{2a}j^{2b}.
466: \end{eqnarray}
467: For $a+b=h-(2n+2)$, we have
468: \begin{equation}
469: E_h^{n+1}(a,b,n+1)=D_h^{n}(a,b+1)=C_h(a+n,b+n+1)>0 \ .
470: \end{equation}
471: {\rm (2)} $i\leq n$ and $j>n+1$, using the induction hypothesis, we
472: derive
473: \begin{eqnarray*}
474: m_h^{n+1}(i,j)&=&m_h^n(i,j)-\frac{j\prod_{r=1}^n (j\triangle
475: r)}{(2n+1)!}m_h^n(i,n+1)\\
476: &=&-(2i-1)!\, j\, \prod_{r=1}^n(j\triangle r) \ \times \\
477: && \left(\sum_{ a+b\leq h-(n+i) }E_h^{n}(a,b,i)i^{2a}j^{2b}-\sum_{
478: a+b\leq h-(n+i)}E_h^{n}(a,b,i)i^{2a}(n+1)^{2b}\right) \ .
479: \end{eqnarray*}
480: Lemma~\ref{L:polynomial} implies
481: \begin{eqnarray*}
482: m_h^{n+1}(i,j)&=&-(2i-1)!\, j\, \prod_{r=1}^{n+1}(j\triangle
483: r)\sum_{a+b\leq h-(n+1+i)}E_h^{n+1}(a,b,i)i^{2a}j^{2b}.
484: \end{eqnarray*}
485: For $h\leq n+i$, we observe $m_h^{n+1}(i,j)=0$ and for $a+b=h-(n+1+i)$,
486: \begin{equation}
487: E_h^{n+1}(a,b,i)=E_h^n(a,b+1,i)=C_h(a+i-1,b+n+1)>0 \ .
488: \end{equation}
489: Accordingly, we have proved
490: \begin{equation}
491: \forall\; i\leq n+1< j;\quad
492: m_h^{n+1}(i,j)=-(2i-1)!j\prod_{r=1}^{n+1}(j\triangle r)\sum_{
493: a+b\leq h-(n+1+i)}E_h^{n+1}(a,b,i)i^{2a}j^{2b} \ .
494: \end{equation}
495: Furthermore
496: \begin{eqnarray}
497: a+b=h-(n+1+i) & \Longrightarrow & E_h^{n+1}(a,b,i)=C_h(a+i-1,b+n+1)>0 \\
498: h<n+1+i & \Longrightarrow & m_h^{n+1}(i,j)=0 \ .
499: \end{eqnarray}
500: Whence assertion (a) holds by induction for any $1\le t< k-1$. We
501: next suppose assertion (b) is true for $t=n$ and consider the case
502: $t=n+1$, i.e., $i>n+1$ and $j>n+1$. First the $i$-loop is executed
503: and produces
504: $$
505: \tilde{m}_h^{n+1}(i,j) = m_h^n(i,j)-\frac{i\,\prod_{r=1}^n(i\triangle r)
506: }{(2n+1)!}m_h^n(n+1,j) \ .
507: $$
508: Secondly the $j$-loop yields
509: \begin{eqnarray}
510: m_h^{n+1}(i,j) & = & \tilde{m}_h^{n+1}(i,j)-\frac{j\prod_{r=1}^n(j\triangle r)}
511: {(2n+1)!}\tilde{m}_h^{n+1}(i,n+1) \ .
512: \end{eqnarray}
513: We accordingly compute
514: \begin{eqnarray*}
515: m_h^{n+1}(i,j) & = & m_h^n(i,j)-\frac{i\,\prod_{r=1}^n(i\triangle r)
516: }{(2n+1)!}m_h^n(n+1,j)\\
517: &&-\frac{j\prod_{r=1}^n(j\triangle r)}{(2n+1)!}
518: \left(m_h^n(i,n+1)-\frac{i\prod_{r=1}^n(i\triangle r)}
519: {(2n+1)!}m_h^n(n+1,n+1)\right)
520: \end{eqnarray*}
521: from which we immediately observe that $m_h^{n+1}(i,j)=m_h^{n+1}(j,i)$ holds.
522: Furthermore
523: \begin{eqnarray*}
524: m_h^{n+1}(i,j)&=&-i\ \left[\prod_{r=1}^n(i\triangle r)\right]\,j\,
525: \left[\prod_{r=1}^n(j\triangle r)\right]\ \times\\
526: &&\sum_{ a+b\leq h-(2n+1)}D_h^{n}(a,b)
527: (i^{2a}j^{2b}-(n+1)^{2a}j^{2b}-i^{2a}(n+1)^{2b}+(n+1)^{2a}(n+1)^{2b})\\
528: &=&-i\ \left[\prod_{r=1}^{n+1}(i\triangle r)\right]\,j\,
529: \left[\prod_{r=1}^{n+1}(j\triangle r)\right]
530: \sum_{ a+b\leq h-(2(n+1)+1)}D_h^{n+1}(a,b)i^{2a}j^{2b}.\\
531: \end{eqnarray*}
532: In particular,
533: \begin{eqnarray*}
534: a+b=h-(2(n+1)+1) &\Rightarrow &
535: D_h^{n+1}(a,b)=D_h^n(a+1,b+1)=C_h(a+n+1,b+n+1)>0 \\
536: h=2n+1 \text{ or } h=2n+2 &\Rightarrow & m_h^{n+1}(i,j)=0 \ .
537: \end{eqnarray*}
538: Thus $m_h^{n+1}(i,j)$ satisfies (b) for any $1\le t< k-1$.
539: \end{proof}
540: We proceed by analyzing the Laurent series
541: \begin{equation}
542: a_{i,j}(z) = \sum_{h\geq 0} m_h^{k-2}(i,j)\frac{(-1)^h}{16^hh!}\, z^{-h}\ .
543: \end{equation}
544: %%%
545: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
546: %%%
547: \begin{lemma}\label{L:asymptotics}
548: \begin{eqnarray}
549: a_{i,j}(z) & = &
550: (-1)^{i+j}\frac{2\Gamma(j+i-\frac{1}{2})}{\sqrt{\pi}}
551: z^{-(j+i-1)}(1+{O}(|z|^{-1})) \ .
552: \end{eqnarray}
553: \end{lemma}
554: %%%
555: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
556: %%%
557: \begin{proof} We shall prove the lemma distinguishing the cases $i<j$ and
558: $i=j$. The former implies by symmetry the case $i>j$. Suppose first
559: $i<j$. By construction of {\bf A}, we have
560: \begin{equation}
561: m_h^{k-2}(i,j)=m_h^{j-1}(i,j)
562: \end{equation}
563: since after the $(j-1)$th step, $m_h^{j-1}(i,j)$ remains unchanged. Consequently
564: we can write $a_{i,j}(z)$ as
565: \begin{equation}
566: a_{i,j}(z)
567: =\sum_{0\le h\le i+j-1}\frac{(-1)^h}{16^h\,h!}\, m_h^{j-1}(i,j)\,z^{-h}
568: + \sum_{ i+j-1<h}\frac{(-1)^h}{16^h\,h!}\, m_h^{j-1}(i,j)\,z^{-h} \ .
569: \end{equation}
570: We consider the terms $m_h^{j-1}(i,j)$ for
571: $0\le h\le j+i-1$. According to Lemma~\ref{L:induction}
572: \begin{equation*}
573: m_h^{j-1}(i,j)=-(2i-1)!\, j\, \prod_{r=1}^{j-1}(j\triangle r)\, \sum_{0\leq
574: a+b\leq h-(j-1+i)}E_h^{j-1}(a,b,i)\, i^{2a}j^{2b}
575: \end{equation*}
576: holds. In particular,
577: \begin{equation}
578: h<j-1+i \quad \Longrightarrow \quad m_h^{j-1}(i,j)=0 \ .
579: \end{equation}
580: Accordingly, the only nonzero coefficient of
581: $\sum_{0\le h\le i+j-1}\frac{(-1)^h}{16^h\,h!}\, m_h^{j-1}(i,j)\,z^{-h}$
582: has index $h=j-1+i$ in which case
583: $$
584: a+b=0\quad \text{\rm and}\quad E_{j-1+i}^{j-1}(0,0,i)=C_{j-1+i}(i-1,j-1)
585: $$
586: holds, i.e.
587: \begin{equation}
588: a_{i,j}(z)= \frac{(-1)^{j+i}C_{j-1+i}(i-1,j-1)(2j-1)!(2i-1)!}
589: {16^{j-1+i}(j-1+i)!}z^{-(j-1+i)}\, (1+{O}(|z|^{-1})) \ .
590: \end{equation}
591: Secondly suppose $i=j$. Then, by definition of ${\bf A}$, the Laurent
592: series $a_{i,i}(z)$ is obtained for $t=i-1$, i.e.~we have
593: \begin{equation}
594: a_{i,i}(z)
595: =\sum_{0\le h\le 2i-1}\frac{(-1)^h}{16^h\,h!}\, m_h^{i-1}(i,i)\,z^{-h}
596: + \sum_{ 2i-1<h}\frac{(-1)^h}{16^h\,h!}\, m_h^{i-1}(i,i)\,z^{-h} \ .
597: \end{equation}
598: Lemma~\ref{L:induction} (b) implies
599: \begin{equation*}
600: m_h^{i-1}(i,i)=-((2i-1)!)^2\sum_{0\leq a+b \leq
601: h-(2i-1)}D_h^{i-1}(a,b)\, i^{2a}i^{2b}.
602: \end{equation*}
603: In particular for $h<2i-1$ we have $m_h^{i-1}(i,i)=0$, thus for
604: $\sum_{0\le h\le 2i-1}\frac{(-1)^h}{16^h\,h!}\, m_h^{i-1}(i,i)\,z^{-h}$ only
605: the index $h=2i-1$ has a nonzero coefficient in which case
606: $$
607: a=b=0 \quad \text{\rm and} \quad D_{2i-1}^{i-1}(0,0)=C_{2i-1}(i-1,i-1)
608: $$
609: holds. We therefore derive
610: \begin{equation}
611: a_{i,i}(z)= \frac{(-1)^{2i}((2i-1)!)^2C_{2i-1}(i-1,i-1)}
612: {16^{2i-1}(2i-1)!}z^{-(2i-1)}(1+{O}(|z|^{-1})).
613: \end{equation}
614: Thus we have proved that we have for $i\leq j$
615: \begin{equation}
616: a_{i,j}(z)= \frac{(-1)^{j+i}C_{j-1+i}(i-1,j-1)(2j-1)!(2i-1)!}
617: {16^{j-1+i}(j-1+i)!}z^{-(j-1+i)}(1+{O}(|z|^{-1})) \ .
618: \end{equation}
619: %%%
620: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
621: %%%
622: {\it Claim $1$.}
623: \begin{equation}
624: C_{j-1+i}(i-1,j-1)=
625: \frac{j\Gamma(2i+2j-1)}{\Gamma(2j+1)\Gamma(2i)} 4^{j+i}.
626: \end{equation}
627: %%%
628: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
629: %%%
630: According to eq.~(\ref{E:qw})
631: \begin{eqnarray*}
632: m_h(i,j)&=& -ij\sum_{0\leq a+b\leq h-1}C_h(a,b)i^{2a}j^{2b}
633: \end{eqnarray*}
634: from which we can conclude for $a+b=h-1$ and $l_m=\min\{a,b\}$
635: \begin{align*}
636: C_h(a,b)&=
637: \sum_{s=0}^{l_m}\binom{h}{a-s}\binom{h-a+s}{b-s}4^{a-s}4^{b-s}
638: 8^{2s+1}2\nonumber\\
639: &=4^{h+1}\sum_{s=0}^{l_m}\binom{h}{a-s}\binom{h-a+s}{2s+1}4^s \ .
640: \end{align*}
641: Therefore
642: \begin{equation}
643: C_{j-1+i}(i-1,j-1)=
644: \frac{j\Gamma(2i+2j-1)}{\Gamma(2j+1)\Gamma(2i)} 4^{j+i} \
645: \end{equation}
646: and Claim $1$ follows. Since $\det[a_{i,j}(z)]_{i,j=1}^{k-1}$ is
647: symmetric, we arrive at
648: \begin{equation*}
649: a_{i,j}(z)=(-1)^{i+j}\frac{2\Gamma(j+i-\frac{1}{2})}
650: {\sqrt{\pi}}z^{-(j+i-1)}(1+{O}(|z|^{-1}))
651: \end{equation*}
652: for any $1\le i,j\le k-1$ and the lemma follows.
653: \end{proof}
654:
655: %%%
656: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
657: %%%
658:
659: {\bf Proof of Theorem~\ref{T:A}.}\\
660: Let
661: \begin{equation}
662: b_{i,j}(z)=(-1)^{i+j}\frac{2\Gamma(j+i-\frac{1}{2})}{\sqrt{\pi}}z^{-(j+i-1)}
663: \ .
664: \end{equation}
665: According to Lemma~\ref{L:asymptotics} we have $a_{i,j}(z)=
666: b_{i,j}(z)\, [1+{O}(|z|^{-1})]$ and we immediately obtain
667: \begin{eqnarray*}
668: \det[a_{i,j}(z)]_{i,j=1}^{k-1} &=&\sum_{\sigma\in S_{k-1}}{\rm
669: sign}(\sigma)\, \prod_{j=1}^{k-1}\left[
670: b_{j,\sigma(j)}(z)\,\left[1+{O}(|z|^{-1})\right]\right]
671: \end{eqnarray*}
672: where $S_{k-1}$ denotes the symmetric group over $k-1$ letters. Furthermore
673: we observe
674: \begin{eqnarray*}
675: \det[b_{i,j}(z)]_{i,j=1}^{k-1}&=&\sum_{\sigma\in S_{k-1}}
676: \text{\rm sign}(\sigma)\,\prod_{j=1}^{k-1}b_{j,\sigma(j)}(z)\\
677: &=&\sum_{\sigma\in S_{k-1}}
678: \text{\rm sign}(\sigma)\,(-1)^{\sum_{j=1}^{k-1}(j+\sigma(j))}\,
679: \left(\frac{2}{\sqrt{\pi}}\right)^{k-1} \times \\
680: & & \qquad \qquad z^{-\sum_{j=1}^{k-1}(j+\sigma(j)-1)}
681: \prod_{j=1}^{k-1}\Gamma(j+\sigma(j)-\frac{1}{2})\\
682: \end{eqnarray*}
683: Since $\sum_{j=1}^{k-1}(j+\sigma(j))=k(k-1)$ we arrive at
684: \begin{eqnarray*}
685: \det[b_{i,j}(z)]_{i,j=1}^{k-1}
686: & = & \left(\frac{2}{\sqrt{\pi}}\right)^{k-1} z^{-(k-1)^2}\,
687: \det\left[\Gamma(j+i-\frac{1}{2})\right]_{i,j=1}^{k-1} \\
688: \end{eqnarray*}
689: and consequently
690: \begin{eqnarray*}
691: \det[a_{i,j}(z)]_{i,j=1}^{k-1} & = & \sum_{\sigma\in S_{k-1}}{\rm
692: sign}(\sigma)\, \prod_{j=1}^{k-1}\left[
693: b_{j,\sigma(j)}(z)\,\left[1+{O}(|z|^{-1})\right]\right]\\
694: & =
695: & \det[\Gamma(j+i-\frac{1}{2})]_{i,j=1}^{k-1}
696: \left(\frac{2}{\sqrt{\pi}}\right)^{k-1}
697: z^{-(k-1)^2}(1+{O}(|z|^{-1})) \ .
698: \end{eqnarray*}
699: We proceed by computing the determinant
700: \begin{equation}
701: \det[\Gamma(j+i-\frac{1}{2})]_{i,j=1}^{k-1} =
702: \prod_{i=1}^{k-1}\Gamma(i+1-\frac{1}{2}) \, \prod_{r=1}^{k-2}r! \ .
703: \end{equation}
704: Since $\Gamma(i+j+1-1/2)=(i+j-1/2)\,\Gamma(i+j-1/2)$, we have for $j>1$
705: \begin{equation*}
706: \Gamma(i+j-\frac{1}{2})=\prod_{r=1}^{j-1}(i+r-\frac{1}{2}) \,
707: \Gamma(i+1-\frac{1}{2}) \ .
708: \end{equation*}
709: We set
710: \begin{equation*}
711: u_{i,j}=\begin{cases} \prod\limits_{r=1}^{j-1}(i+r-\frac{1}{2})& j>1\\
712: 1& j=1\\
713: \end{cases}
714: \end{equation*}
715: and compute
716: \begin{align*}
717: \det[\Gamma(j+i-\frac{1}{2})]_{i,j=1}^{k-1}&=
718: \prod_{i=1}^{k-1}\Gamma(i+1-\frac{1}{2})
719: \det[u_{i,j}]_{i,j=1}^{k-1}
720: =\prod_{i=1}^{k-1}\Gamma(i+1-\frac{1}{2})\det[i^{j-1}]_{i,j=1}^{k-1}.
721: \end{align*}
722: The determinant $\det[i^{j-1}]_{i,j=1}^{k-1}$ is a Vandermonde
723: determinant, whence
724: \begin{equation*}
725: \det[i^{j-1}]_{i,j=1}^{k-1}=\sum_{1\leq i_1<i_2\leq
726: k-1}(i_2-i_1)=\prod_{r=1}^{k-2}r! \ .
727: \end{equation*}
728: Therefore we have shown
729: \begin{eqnarray}\label{E:www}
730: \det[a_{i,j}(z)]_{i,j=1}^{k-1} & = &
731: \left[\prod_{i=1}^{k-1}\Gamma(i+1-\frac{1}{2}) \, \prod_{r=1}^{k-2}r!\right]\,
732: \left(\frac{2}{\sqrt{\pi}}\right)^{k-1}
733: z^{-(k-1)^2}(1+{O}(|z|^{-1})) \ .
734: \end{eqnarray}
735: It remains to combine our findings:
736: the approximation of the Bessel
737: function eq.~(\ref{E:bessel2}) and eq.~(\ref{E:m}) imply for
738: $-\frac{\pi}{2}< \text{\rm arg(z)} < \frac{\pi}{2}$
739: \begin{equation*}
740: I_{i-j}(2z)-I_{i+j}(2z)=\frac{e^{2z}}{2\sqrt{\pi
741: z}}\left(\sum_{h=1}^H m_h(i,j)\frac{(-1)^h}{16^hh!}\,
742: z^{-h}+O(|z|^{-H-1})\right) \ .
743: \end{equation*}
744: Let
745: \begin{equation}
746: e_{i,j}^H(z)=\sum_{h=1}^H m_h(i,j)\frac{(-1)^h}{16^hh!}\,z^{-h}
747: \end{equation}
748: then we have
749: \begin{equation}
750: F_k(z)=\det[I_{i-j}(2z)-I_{i+j}(2z)]_{i,j=1}^{k-1}
751: =\left(\frac{e^{2z}}{2\sqrt{\pi z}}\right)^{k-1}\,
752: \left[\det\left[e_{i,j}^H\right]_{i,j=1}^{k-1}+O(|z|^{-H-1})\right].
753: \end{equation}
754: Lemma~\ref{L:induction} and Lemma~\ref{L:asymptotics} provide an
755: interpretation of $\det[e_{i,j}^H(z)]_{i,j=1}^{k-1}$: for
756: \begin{equation}
757: H> (k-1)^2
758: \end{equation}
759: we can conclude
760: \begin{equation*}
761: \det[e_{i,j}^H(z)]_{i,j=1}^{k-1}=\det[b_{i,j}(z)]_{i,j=1}^{k-1} \,
762: \left[1+O(\vert z\vert^{-1})\right] \ .
763: \end{equation*}
764: Accordingly we derive
765: \begin{eqnarray*}
766: F_k(z) & = & \left(\frac{e^{2z}}{2\sqrt{\pi z}}\right)^{k-1}\,
767: \det[b_{i,j}(z)]_{i,j=1}^{k-1} \, \left[1+O(\vert
768: z\vert^{-1})\right] \ .
769: \end{eqnarray*}
770: Since
771: $$
772: \det[b_{i,j}(z)]_{i,j=1}^{k-1} =
773: \left(\frac{2}{\sqrt{\pi}}\right)^{k-1}
774: z^{-(k-1)^2}\,\det[\Gamma(j+i-\frac{1}{2})]_{i,j=1}^{k-1}
775: $$
776: and $F_k(z)$ is an even function, we obtain for $\text{\rm arg}(z)\ne
777: \pm \frac{\pi}{2}$
778: \begin{eqnarray}\label{E:das}
779: F_k(z) & = &
780: \left[\prod_{i=1}^{k-1}\Gamma(i+1-\frac{1}{2}) \, \prod_{r=1}^{k-2}r!\right]
781: \left(\frac{e^{2z}}{\pi}\right)^{k-1}\, z^{-(k-1)^2-\frac{k-1}{2}}\;
782: \ (1+{O}(|z|^{-1}))
783: \end{eqnarray}
784: and the proof of the theorem is complete. \hfill $\square$
785:
786:
787:
788: %%%
789: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
790: %%%
791: \section{Proof of Theorem~\ref{T:B}}
792:
793: %%%
794: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
795: %%%
796:
797: Suppose $k=4m$, $m\in\mathbb{N}$,
798: $p=(k-1)^2+\frac{k-2}{2}=(4m-1)^2+2m-1$ and
799: %$(k-1)^2+\frac{k-1}{2}=(4m-1)^2+2m-1+\frac{1}{2}$.
800: \begin{eqnarray} \label{E:case1}
801: \quad g_k(z) & = & \tilde{c}_{k}\,
802: \left[I_0((2k-2)z)\,z^{-p}-\sum_{j=1}^{p}a_{k,j}\,z^{-j}\right], \
803: \text{\rm where} \ a_{k,j}=[z^{p-j}]\,I_0((2k-2)z).
804: \end{eqnarray}
805: For $k=4m+2$, let $p=(k-1)^2+\frac{k-2}{2}=(4m+1)^2+2m$ and
806: \begin{eqnarray} \label{E:case3}
807: \quad g_k(z) &= &
808: \tilde{c}_{k}\left[I_1((2k-2)z)\,z^{-p}-\sum_{j=1}^{p}a_{k,j}\,z^{-j}\right],
809: \ \text{\rm where} \ a_{k,j}=[z^{p-j}]\,I_1((2k-2)z).
810: \end{eqnarray}
811: For $k=4m+1$, let $p=(k-1)^2+\frac{k-1}{2}=(4m)^2+2m$ we set
812: \begin{eqnarray} \label{E:case2}
813: \quad g_k(z) & = & \tilde{c}_{k}\,
814: \left[\cosh((2k-2)z)\,z^{-p}-\sum_{j=1}^{p}a_{k,j}\,z^{-j}\right],\
815: \text{\rm where} \ a_{k,j}=[z^{p-j}]\,\cosh((2k-2)z).
816: \end{eqnarray}
817: Finally, for $k=4m+3$, let $p=(k-1)^2+\frac{k-1}{2}=(4m+2)^2+2m+1$
818: and
819: \begin{eqnarray} \label{E:case4}
820: \quad g_k(z) & = &
821: \tilde{c}_{k}\left[\sinh((2k-2)z)\,z^{-p}-\sum_{j=1}^{p}a_{k,j}\,z^{-j}\right],
822: \ \text{\rm where} \ \ a_{k,j}=[z^{p-j}]\,\sinh((2k-2)z).
823: \end{eqnarray}
824: The functions given in eq.~(\ref{E:case1})-(\ref{E:case4}) are
825: entire, even and the constants $\tilde{c}_{k}$ satisfy
826: \begin{equation*} g_k(|z|)\sim
827: c_k'\,e^{(2k-2)|z|}\,|z|^{-(k-1)^2-\frac{k-1}{2}},\quad \ \text{\rm
828: as} \ \vert z\vert\to \infty \
829: \end{equation*}
830: where $c_k'=\pi^{-(k-1)}\prod_{i=1}^{k-1}\Gamma(i+1-\frac{1}{2}) \,
831: \prod_{r=1}^{k-2}r!$.
832: \begin{proof}
833: %%%
834: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
835: %%%
836: {\it Claim $1$.} Suppose $z\in\mathbb{C}\setminus\mathbb{R}$, then
837: we have
838: \begin{equation}\label{E:F_kpro}
839: |F_k(z)|={o}(|z|^{-1}F_k(|z|)).
840: \end{equation}
841: %%%
842: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
843: %%%
844: To prove Claim $1$, we conclude from Theorem~\ref{T:A} that
845: \begin{equation}\label{E:F_kasy}
846: F_k(z)=
847: c_k'\,e^{(2k-2)z}\,z^{-(k-1)^2-\frac{k-1}{2}}(1+{O}(|z|^{-1}))\quad\text{\rm
848: for}\ \arg(z)\neq \pm \pi/2\ ,
849: \end{equation}
850: where $c_k'=\pi^{-(k-1)}\prod_{i=1}^{k-1}\Gamma(i+1-\frac{1}{2}) \,
851: \prod_{r=1}^{k-2}r!$ holds. We write $z=re^{i\theta}$ and obtain for
852: $\theta\neq 0,\ \pi,\ \pm \pi/2$
853: \begin{eqnarray}\label{E:fast}
854: \frac{|F_k(z)|}{|z|^{-1}F_k(|z|)} & = & e^{-2(k-1)(1-\cos\theta)\,
855: r}\ r \ \left({O}(1)+{O}(|z|^{-1})\right).
856: \end{eqnarray}
857: Therefore we have $|F_k(z)|={o}( |z|^{-1}F_k(|z|))$ for
858: $\arg(z)\neq0,\pi, \pm \pi/2$. Since $|F_k(z)|$ and $|z|^{-1}F_k(|z|)$
859: are continuous, eq.~(\ref{E:fast}) implies
860: \begin{eqnarray}
861: |F_k(z)| & = & {o}(|z|^{-1}F_k(|z|)), \quad \text{\rm for
862: $z\in\mathbb{C} \setminus\mathbb{R}$}.
863: \end{eqnarray}
864: whence Claim $1$.\\
865: %%%
866: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
867: %%%
868: {\it Claim 2.} For any $k\ge 2$, the functions given in
869: eq.~(\ref{E:case1})-(\ref{E:case4}) satisfy
870: \begin{equation}\label{E:g_kpro}
871: |g_k(z)|={o}(|z|^{-1}g_k(|z|)) \quad \text{\rm for $z\in\mathbb{C}
872: \setminus\mathbb{R}$}
873: \end{equation}
874: and
875: \begin{equation*} g_k(|z|)=
876: c_k'\,e^{(2k-2)|z|}|z|^{-(k-1)^2-\frac{k-1}{2}}(1+{O}(|z|^{-1})).
877: \end{equation*}
878: %%%
879: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
880: %%%
881: Suppose first $k=4m$ or $4m+2$. Then we have
882: \begin{equation}\label{E:g_keven}
883: g_k(z)=\tilde{c}_{k}
884: \left(I_s((2k-2)z)z^{-p}-\sum_{j=1}^{p}a_{k,j}z^{-j}\right),\quad
885: s=0\ \text{or}\ 1,
886: \end{equation}
887: where $p=(k-1)^2+\frac{k-2}{2}$. For
888: $-\frac{\pi}{2}<\text{arg}(z)<\frac{\pi}{2}$, we have
889: \begin{equation}\label{E:bessel3}
890: I_s(z)=\frac{e^{z}}{\sqrt{2 \pi
891: z}}\left(\sum_{h=0}^{H}\frac{(-1)^{h}}{h!8^h}\prod\limits_{t=1}^h
892: (4s^2-(2t-1)^2)z^{-h}+{O}(|z|^{-H-1})\right).
893: \end{equation}
894: Using eq.(\ref{E:bessel3}) we derive for sufficiently large $\vert
895: z\vert$
896: \begin{align*}
897: \frac{|g_k(z)|}{|z|^{-1}g_k(|z|)}&\leq
898: \frac{|I_s((2k-2)z)||z|^{-p}+\sum_{j=1}^{p}a_{k,j}|z|^{-j}}
899: {I_s((2k-2)|z|)|z|^{-p-1}-\sum_{j=1}^{p}a_{k,j}|z|^{-j-1}} \\
900: & \le C_0\; e^{-2(k-1)(1-\cos\theta)r}\; r\; \ ,
901: \end{align*}
902: where $C_0>0$ is some constant. Since $g_k(z)$ is even we have shown
903:
904: \begin{equation}\label{E:fast2}
905: |g_k(z)| ={o}(|z|^{-1}g_k(|z|)) \quad \; \text{\rm where
906: }\arg(z)\not\in\{0,\pi,\frac{\pi}{2},-\frac{\pi}{2} \}.
907: \end{equation}
908: Since $g_k(z)$ is continuous eq.~(\ref{E:fast2}) implies
909: $|g_k(z)|={o}(|z|^{-1}g_k(|z|)) \quad \text{\rm for $z\in\mathbb{C}
910: \setminus\mathbb{R}$}$ .\\
911: By eq.~(\ref{E:bessel3}) and the definition of $g_k(z)$, we can
912: obtain that
913: \begin{align*}
914: g_k(|z|)&=\tilde{c}_{k}
915: \left(I_s((2k-2)|z|)|z|^{-p}-\sum_{j=1}^{p}a_{k,j}|z|^{-j}\right)\\
916: &=\tilde{c}_{k}\frac{e^{(2k-2)|z|}}{2\sqrt{(k-1)\pi}\,|z|^{p+\frac{1}{2}}}
917: (1+O(|z|^{-1}))-\tilde{c}_{k}\sum_{j=1}^{p}a_{k,j}|z|^{-j}\\
918: &= c_k'\,e^{(2k-2)|z|}|z|^{-(k-1)^2-\frac{k-1}{2}}(1+{O}(|z|^{-1})).
919: \end{align*}
920:
921:
922: For $k=4m+1$ or $4m+3$, $g_k(z)$ satisfies
923: \begin{align*}
924: |g_k(z)|&\leq \tilde{c}_k\left(\frac{|e^{(2k-2)z}|+|e^{-(2k-2)z}|}
925: {2}|z|^{-p}+\sum_{j=1}^{p}a_{k,j}|z|^{-j}\right)\\
926: &=\tilde{c}_k\left(\frac{e^{(2k-2)r\cos\theta}+
927: e^{-(2k-2)r\cos\theta}}{2}r^{-p}+\sum_{j=1}^{p}a_{k,j}r^{-j}\right)
928: \end{align*}
929: where $p=(k-1)^2+\frac{k-1}{2}$ and consequently for sufficiently
930: large $\vert z\vert$
931: \begin{equation}\label{E:oha}
932: \frac{|g_k(z)|}{|z|^{-1}g_k(|z|)} \le C_1\; r\;
933: e^{-(2k-2)r(1-|\cos\theta|)}
934: \end{equation}
935: for some $C_1>0$. eq.~(\ref{E:oha}) shows
936: \begin{equation}
937: \forall z\in\mathbb{C}\setminus \mathbb{R} \qquad
938: |g_k(z)|={o}(|z|^{-1}g_k(|z|)) \ .
939: \end{equation}
940: For $k=4m+1$ we derive
941: \begin{align*}
942: g_k(|z|)&=\tilde{c}_k\left(\cosh((2k-2)|z|)|z|^{-p}
943: -\sum_{j=1}^{p}a_{k,j}|z|^{-j}\right)\\
944: &=\tilde{c}_k\left(\frac{e^{(2k-2)|z|}+e^{-(2k-2)|z|}}{2}|z|^{-p}
945: -\sum_{j=1}^{p}a_{k,j}|z|^{-j}\right)\\
946: &=c_k'\,e^{(2k-2)|z|}|z|^{-(k-1)^2-\frac{k-1}{2}}(1+{O}(|z|^{-1})).
947: \end{align*}
948: The case $k=4m+3$ follows analogously.
949: We can conclude from
950: \begin{equation*}
951: F_k(z) =
952: c_k'\,e^{(2k-2)z}z^{-(k-1)^2-\frac{k-1}{2}}(1+{O}(|z|^{-1}))\quad\text{\rm
953: for}\ \arg(z)\neq \pm \pi/2, \\
954: \end{equation*}
955: and
956: \begin{equation*} g_k(|z|)=
957: c_k'\,e^{(2k-2)|z|}|z|^{-(k-1)^2-\frac{k-1}{2}}(1+{O}(|z|^{-1})).
958: \end{equation*}
959: that $F_k(\vert z\vert) = g_k(\vert z\vert)(1+{O}(|z|^{-1}))$
960: holds.
961: %%%
962: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
963: %%%
964: To summarize we have shown
965: \begin{eqnarray*}
966: |F_k(z)| & = & {o}(|z|^{-1}F_k(|z|)) \quad\text{\rm for
967: $z\in\mathbb{C}\setminus \mathbb{R}$}\\
968: |g_k(z)| & = & {o}(|z|^{-1}g_k(|z|)) \quad\text{\rm for
969: $z\in\mathbb{C}\setminus \mathbb{R}$}\\
970: F_k(\vert z\vert)& = & g_k(\vert z\vert)(1+{O}(|z|^{-1})).
971: \end{eqnarray*}
972: We can accordingly conclude that
973: \begin{equation}\label{E:wichtig}
974: |F_k(z)-g_k(z)|={O}(|z|^{-1}g_k(|z|)),
975: \end{equation}
976: uniformly for all $z$ with $|z|\geq 1$.\\
977: %%%
978: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
979: %%%
980: {\it Claim 3.} For arbitrary $k\ge 2$ we have
981: \begin{equation}\label{E:coeff33}
982: f_k(n)\sim c_k\, n^{-(k-1)^2-\frac{k-1}{2}}\;(2k-2)^{2n}
983: \quad\text{\rm where $c_k>0$} \ .
984: \end{equation}
985: %%%
986: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
987: %%%
988: To prove Claim $3$ we compute, using eq.~(\ref{E:wichtig})
989: \begin{eqnarray*}
990: \left| [z^{2n}]\,(F_k(z)-g_k(z))\right| &\leq&\int_{\vert
991: z\vert=\frac{n}{k-1}}\frac{\left|F_k(z)-g_k(z)\right|}{|z|^{2n+1}}|\mathrm{d}z|\\
992: &\leq& c\int_{\vert
993: z\vert=\frac{n}{k-1}}\frac{|z|^{-1}g_k(|z|)}{|z|^{2n+1}}|\mathrm{d}z|,\\
994: \end{eqnarray*}
995: where $c$ is a positive constant. For $k=4m\ \text{or}\ 4m+2$ we
996: have $p=(k-1)^2+\frac{k-2}{2}$ and substituting for $g_k(\vert
997: z\vert)$
998: \begin{align*}
999: \left| [z^{2n}]\,(F_k(z)-g_k(z))\right| &\leq c'\int_{\vert
1000: z\vert=\frac{n}{k-1}}\frac{|z|^{-1}|z|^{-p-\frac{1}{2}}
1001: e^{(2k-2)|z|}}{|z|^{2n+1}}|\mathrm{d}z|\\
1002: &=c'\, e^{(2k-2)\cdot\frac{n}{k-1}}\, \left(\frac{n}{k-1}\right)^{
1003: -2n-2-p-\frac{1}{2}}\, 2\pi\frac{n}{k-1}\\
1004: &=c''\, e^{2n}(k-1)^{2n}n^{-(2n+p+\frac{3}{2})}
1005: \end{align*}
1006: where $c',c''$ are positive constants. By definition of the Bessel
1007: function, see eq.~(\ref{E:bessel}),~(\ref{E:g_keven}) and using
1008: Stirling's formula
1009: \begin{equation}\label{stirling1}
1010: \begin{split}
1011: [z^{2n}]\,g_k(z)&=\tilde{c}_{k}\,[z^{2n+p}]\,I_s((2k-2)z)
1012: =\tilde{c}_{k}\,\frac{(k-1)^{2n+p}}{(n+\frac{p-s}{2})!(n+\frac{p+s}{2})!}\\
1013: &\sim \tilde{c}_{k}'\; e^{2n}(k-1)^{2n}n^{-(2n+p+1)}.
1014: \end{split}
1015: \end{equation}
1016: Here $s$ only depends on $k$ and $\tilde{c}_{k}'$ is a positive
1017: constant. Therefore we can conclude
1018: \begin{equation}
1019: [z^{2n}]\,F_k(z) \sim [z^{2n}]\, g_k(z) \ ,
1020: \end{equation}
1021: whence
1022: \begin{eqnarray*}
1023: f_{k}(n) &=& (2n)!\,[z^{2n}]\, F_k(z) \, \sim\,
1024: (2n)!\tilde{c}_{k}\,
1025: \frac{(k-1)^{2n+p}}{(n+\frac{p-s}{2})!(n+\frac{p+s}{2})!}\\
1026: &\sim & c_k\,(2k-2)^{2n}n^{-(k-1)^2-\frac{k-1}{2}} \ .
1027: \end{eqnarray*}
1028: In case of $k=4m+1$ or $4m+3$ we have $p=(k-1)^2+\frac{k-1}{2}$ and
1029: compute
1030: \begin{align*}
1031: \left| [z^{2n}](F_k(z)-g_k(z))\right| & \leq c'\int_{\vert
1032: z\vert=\frac{n}{k-1}}
1033: \frac{|z|^{-1}|z|^{-p}e^{(2k-2)|z|}}{|z|^{2n+1}}|\mathrm{d}z|\\
1034: &=c'e^{(2k-2)\, \frac{n}{k-1}}\,
1035: \left(\frac{n}{k-1}\right)^{-2n-2-p}
1036: \, 2\pi\frac{n}{k-1}\\
1037: &=c''e^{2n}(k-1)^{2n}n^{-(2n+p+1)}
1038: \end{align*}
1039: where $c',c''$ are positive constants. For $k=4m+1$ we obtain
1040: \begin{equation}\label{stirling2}
1041: [z^{2n}]\,g_k(z) = \tilde{c}_{k} \,
1042: [z^{2n+p}]\,\cosh((2k-2)z)= \tilde{c}_{k}\,
1043: \frac{(2k-2)^{2n+p}}{(2n+p)!} \sim \tilde{c}_{k}'\;
1044: e^{2n}(k-1)^{2n}n^{-(2n+p+\frac{1}{2})}
1045: \end{equation}
1046: and for $k=4m+3$
1047: \begin{equation}\label{stirling3}
1048: [z^{2n}]\,g_k(z)=\tilde{c}_{k}\,[z^{2n+p}]\,\sinh((2k-2)z)
1049: =\tilde{c}_{k}\, \frac{(2k-2)^{2n+p}}{(2n+p)!} \sim \tilde{c}_{k}'\;
1050: e^{2n}(k-1)^{2n}n^{-(2n+p+\frac{1}{2})}.
1051: \end{equation}
1052: Since $\left| [z^{2n}]\,(F_k(z)-g_k(z))\right|\le
1053: c''\,e^{2n}(k-1)^{2n}n^{-(2n+p+1)}$ eq.~(\ref{stirling2}) and
1054: (\ref{stirling3}) guarantee
1055: \begin{equation}
1056: [z^{2n}]\,F_k(z) \sim [z^{2n}]\,g_k(z) \ .
1057: \end{equation}
1058: Accordingly we obtain
1059: \begin{equation}
1060: f_{k}(n)=(2n)!\,[z^{2n}]\,F_k(z)\sim (2n)!\,\tilde{c}_{k}\,
1061: \frac{(2k-2)^{2n+p}}{(2n+p)!}\sim c_k\, n^{-(k-1)^2-\frac{k-1}{2}}
1062: \, (2k-2)^{2n}
1063: \end{equation}
1064: and Theorem~\ref{T:B} follows.
1065: \end{proof}
1066:
1067:
1068: %%%
1069: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1070: %%%
1071: {\bf Acknowledgments.}
1072: %%%
1073: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1074: %%%
1075: This work was supported by the 973 Project, the PCSIRT Project of the
1076: Ministry of Education, the Ministry of Science and Technology, and
1077: the National Science Foundation of China.
1078:
1079: \bibliographystyle{amsplain}
1080: \begin{thebibliography}{10}
1081:
1082: \bibitem{NBS:70} M. Abramowitz and I.A. Stegun, eds.
1083: \textit{Handbook of Mathematical Functions with Formulas, Graphs,
1084: and Mathematical Tables}, NBS Applied Mathematics Series 55,
1085: National Bureau of Standards, Washington, DC (1964).
1086:
1087:
1088: \bibitem{CDDSY} W.Y.C. Chen, E.Y.P. Deng, R.R.X. Du, R.P. Stanley and C.H. Yan, {\it Crossings and Nestings of Matchings and Partitions},
1089: Trans. Amer. Math. Soc. {\bf 359}, 1555-1575, (2007).
1090:
1091: \bibitem{vac} W.Y.C. Chen, J. Qin, and C.M. Reidys, {\it Crossings and
1092: Nestings in Tangled Diagrams}, preprint. Available from {\tt
1093: arXiv:0710.4053}.
1094:
1095: \bibitem{IRC}G.P. Egorychev, \textit{Integral Representation and the
1096: computation of combinatorial sums}, Translations of mathematical
1097: monographs American Mathematical Society {\bf 59}.
1098:
1099: \bibitem{Gessel} I.M. Gessel, {\it Symmetric functions and P-recursiveness}
1100: J.Comb. Theory Ser. A. {\bf 53}, 257-285, (1990).
1101:
1102: \bibitem{GZ} I.M. Gessel and D. Zeilberger, {\it Random Walk in a Weyl
1103: chamber}, Proc. Amer. Math. Soc. {\bf 115}, 27-31, (1992).
1104:
1105: \bibitem{GM} D. Grabiner and P. Magyar, {\it Random Walks in a Weyl Chamber
1106: and the decomposition of tensor powers}, J. Alg. Combinatorics {\bf
1107: 2}, 239-260, (1993).
1108:
1109: \bibitem{Henrion:00} D. Henrion and M. Sebek,
1110: \textit{Improved Polynomial Matrix Determinant Computation}, IEEE
1111: Trans. CAS-PT I. Fundam. Theory Appl. 46, 1307-1308, (1999).
1112:
1113: \bibitem{Odlyzko:95a} A.M. Odlyzko,
1114: \textit{Asymptotic enumeration methods}, Handbook of combinatorics
1115: Vol. {\bf 2}, 1021-1231, (1995).
1116:
1117: \bibitem{Regev:81} A. Regev, {\it Asymptotic values for degrees associated
1118: with strips of Young diagrams}, Adv. Math. Vol. {\bf 41}, 115-136,
1119: (1981).
1120:
1121: \bibitem{S} R. Stanley, {\it Differentiably Finite Power Series}, Europ. J.
1122: Combinatorics {\bf 1}, 175-188, (1980).
1123:
1124: \bibitem{Van:79} P.M. Van Dooren, P. Dewilde and J. Vandewalle,
1125: \textit{On the Determination of the Smith-MacMillan Form of a
1126: Rational Matrix From Its Laurent Expansion}, IEEE Trans on Circuits
1127: and Systems, Vol. {\bf 26}, No. 3, 180-189, (1979).
1128:
1129: \bibitem{Z} D. Zeilberger, {\it A Holonomic systems approach to special
1130: functions identities}, J. of Computational and Applied Math. {\bf
1131: 32}, 321-368, (1990).
1132:
1133: \end{thebibliography}
1134:
1135: %%%
1136: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1137: %%%
1138:
1139:
1140: \end{document}
1141:
1142:
1143: The idea is to employ Theorem~\ref{T:A} and the subtraction of
1144: singularity-principle. In view of the validity of the approximation
1145: guaranteed via Theorem~\ref{T:A} for $-\frac{\pi}{2}< \text{arg}(z)
1146: <\frac{\pi}{2}$ and the fact that $F_k(z)$ is even and entire,
1147: only the case $|\text{arg}(z)|=\pi/2$ has to be considered.
1148: As we shall use contour integration along discs of radius
1149: $\frac{n}{k-1}$ in order to compute the coefficient $[z^{2n}]$,
1150: it suffices to prove the following \\
1151: