1:
2: \documentclass[]{article}
3:
4: \input epsf
5:
6: \renewcommand{\theequation}{\arabic{section}.\arabic{equation}}
7:
8: \def\dsp{\displaystyle}
9: \def\pref#1{(\ref{#1})}
10: \def\intp{\int_0^\infty}
11: \def\intr{\int_{-\infty}^\infty}
12: \def\bigO{{\cal O}}
13: \def\C{{\cal C}}
14:
15:
16: \def\FG#1#2#3#4{
17: {}_2F_1\left(
18: \begin{array}{c}
19: \begin{array}{cc} \hskip-10pt#1,{\ } #2 \end{array}\\
20: \begin{array}{c} \hskip-10pt#3 \end{array}
21: \end{array}
22: \hskip-8pt;\,#4
23: \right)}
24:
25: \def\CHG#1#2#3{
26: {}_1F_1\left(
27: \begin{array}{c}
28: \begin{array}{c} \hskip-10pt#1 \end{array}\\
29: \begin{array}{c} \hskip-10pt#2 \end{array}
30: \end{array}
31: \hskip-8pt;\,#3
32: \right)}
33:
34: \def\FFF#1#2#3#4#5#6{
35: {}_3F_2\left(
36: \begin{array}{c}
37: \begin{array}{c}\hskip-10pt#1,#2,#3\end{array}\\
38: \begin{array}{c}\hskip-10pt #4,#5\end{array}
39: \end{array}
40: \hskip-8pt;\,#6
41: \right)}
42:
43: \def\sfrac#1#2{{{\lower.6ex
44: \hbox{$\scriptstyle#1$}}\over
45: {\raise.7ex
46: \hbox{$\scriptstyle#2$}}}}
47:
48: \begin{document}
49: \title{
50: Asymptotics of a ${}_3F_2$ polynomial associated with the
51: Catalan-Larcombe-French sequence}
52:
53: \author{Nico M. Temme\\
54: CWI, P.O. Box 94079, NL-1090 GB Amsterdam, The Netherlands\\
55: e-mail: {\tt Nico.Temme@cwi.nl}
56: }
57:
58:
59: \maketitle
60: \begin{abstract}
61: \noindent
62: The large $n$ behaviour of the hypergeometric polynomial
63: $$
64: \FFF{-n}{\sfrac12}{\sfrac12}{\sfrac12-n}{\sfrac12-n}{-1}
65: $$
66: is considered by using integral representations of this polynomial.
67: This ${}_3F_2$ polynomial is associated with the
68: Catalan-Larcombe-French sequence. Several other representations
69: are mentioned, with references to the literature,
70: and another asymptotic method is described by using
71: a generating function of the sequence. The results are similar to
72: those obtained by Clark (2004) who used a binomial sum for
73: obtaining an asymptotic expansion.
74:
75:
76:
77: \end{abstract}
78:
79: \vskip 0.8cm \noindent
80: Mathematics Subject Classification 2000:
81: %\par\noindent
82: 41A60, 33C20, 11B83, 33C10.
83: \par\noindent
84: Keywords \& Phrases:
85: %\par\noindent
86: Catalan-Larcombe-French sequence,
87: asymptotic expansion,
88: hypergeometric polynomial,
89: modified Bessel function.
90:
91:
92:
93:
94: %
95: %
96: \section{The problem}
97: \label{sec.TP}
98: %
99: Find the large $n$ asymptotics of
100: %
101: \begin{equation}\label{P1}
102: f(n)=
103: \FFF{-n}{\sfrac12}{\sfrac12}{\sfrac12-n}{\sfrac12-n}{-1}
104: \end{equation}
105: %
106: Peter Larcombe conjectured that
107: $\lim_{n\to\infty} f(n)=2$
108: and Tom Koornwinder gave a proof, based on dominated convergence. See
109: for details
110: of the proof
111: \cite{Larcombe:2006:FPL}, where also a different representation of $f(n)$ is
112: considered in the form
113: \begin{equation}\label{P2}
114: f(n)=
115: 2^{n}\FFF{-n}{-\sfrac12n}{\sfrac12-\sfrac12n}{\sfrac12-n}{\sfrac12-n}{1}.
116: \end{equation}
117: The equivalence of these two forms follows from a quadratic
118: transformation of the ${}_{3}F_{2}-$functions as given in \cite[Ex.~4(iv), p.97]{Bailey:1964:GHS}, that is,
119: %
120: \renewcommand{\arraystretch}{1.5}
121: \begin{equation}\label{P3}\begin{array}{l}
122: \FFF{a}{b}{c}{1+a-b}{1+c-c}{z}=\\
123: \quad\quad\quad\quad
124: (1-z)^{-a}\FFF{\sfrac12a}{\sfrac12+\sfrac12a}{1+a-b-c}{1+a-b}{1+a-c}{\dsp{\frac{-4z}{(1-z)^{2}}}}.
125: \end{array}
126: \end{equation}
127: \renewcommand{\arraystretch}{1.0}
128: %
129: with $a=-n$, $b=c=\frac12$, and $z=-1$.
130: %
131: Another form is given by (see \cite[Eq. (A2)]{Larcombe:2004:ANG})
132: %
133: \begin{equation}
134: \label{P4}
135: f(n)=\frac{n!}{2^{n}(\frac12)_{n}}\FFF{-n}{-n}{\frac12}{1}{\frac12-n}{-1}.
136: \end{equation}
137: %
138:
139: In \cite{Clark:2004:AEC} an asymptotic expansion of $\frac12f(n)$ has been
140: derived. The asymptotic analysis is based on the representation
141: %
142: \begin{equation}
143: \label{P5}
144: P_{n}=\frac{1}{n!}\sum_{p+q=n}{2n\choose p}
145: {2q\choose q}\frac{(2p)!\,(2q)!}{p!\,q!}.
146: \end{equation}
147: %
148: By using the relation
149: %
150: \begin{equation}
151: \label{P6}
152: (2n)!=2^{2n}n!(\sfrac12)_{n},\quad n=0,1,2,\ldots,\end{equation}
153: %
154: it is straightforward to verify that \pref{P5} can be written
155: as
156: %
157: \begin{equation}
158: \label{P7}
159: P_{n}=\frac{2^{4n}}{n!}\sum_{p=0}^{n}\frac{(\frac12)_{p}(\frac12)_{p}(\frac12)_{n-p}(\frac12)_{n-p}}{p!\,(n-p)!}.
160: \end{equation}
161: %
162: By using
163: %
164: \begin{equation}
165: \label{P8}
166: (a)_{n-k}=(-1)^{k}\frac{(a)_{n}}{(1-a-n)_{k}},
167: \end{equation}
168: %
169: it follows that
170: %
171: \begin{equation}
172: \label{P9}
173: P_{n}=\frac{2^{4n}(\frac12)_{n}(\frac12)_{n}}{n!\,n!}
174: \sum_{p=0}^{n}(-1)^{p}\frac{(-n)_{p}(\frac12)_{p}(\frac12)_{p}}{p!\,(\frac12-n)_{p}(\frac12-n)_{p}},
175: \end{equation}
176: %
177: that is,
178: %
179: \begin{equation}
180: \label{P10}
181: P_{n}=\frac{2^{4n}(\frac12)_{n}(\frac12)_{n}}{n!\,n!}
182: \FFF{-n}{\sfrac12}{\sfrac12}{\sfrac12-n}{\sfrac12-n}{-1},
183: \end{equation}
184: %
185: which gives the relation with $f(n)$ by using \pref{P1}:
186: %
187: \begin{equation}
188: \label{P11}
189: P_{n}=\frac{2^{4n}(\frac12)_{n}(\frac12)_{n}}{n!\,n!}\,f(n)={2n
190: \choose n}^{2}\,f(n).
191: \end{equation}
192: %
193:
194: The numbers $P_{n}$ are for $n=0,1,2,,\ldots$ known as the elements of the
195: sequence (A053175) $\{1,8,80, 896,10816,\ldots\}$, called the {\em
196: Catalan-Larcombe-French} sequence, which is originally discussed by
197: Catalan \cite{Catalan:1887:SNS}. See the
198: {\em On-Line Encyclopedia of Integer Sequences}
199: http://www.research.att.com/~njas/sequences/ .
200:
201: In this paper we derive a complete asymptotic expansion of the
202: numbers $P_{n}$ by using integral representations of the
203: corresponding ${}_{3}F_{2}-$functions. Our results are the same as
204: those obtained by Clark \cite{Clark:2004:AEC}, who used the binomial
205: sum in \pref{P5} without reference to the ${}_{3}F_{2}-$functions.
206: %
207:
208:
209: %
210: \section{Transformations}
211: \label{sec.TR}
212: %
213: We derive an integral representation of the ${}_{3}F_{2}-$function of
214: \pref{P1} by using several transformations for special functions. We
215: start with the beta integral
216: %
217: \begin{equation}\label{T0}
218: B(x,y)=\frac{\Gamma(x)\Gamma(y)}{\Gamma(x+y)}=\int_{0}^{1}t^{x-1}(1-t)^{y-1}\,dt
219: \end{equation}
220: %
221: and use it in the form
222: \begin{equation}\label{T1}
223: \frac{\left(\sfrac12\right)_k}{\left(\sfrac12-n\right)_k}=
224: \frac{(-1)^kn!}{\sqrt{{\pi}}\Gamma(n+\sfrac12)}\int_0^1
225: t^{k-\frac12}(1-t)^{n-k-\frac12}\,dt,\quad k=0,1,\ldots, n.
226: \end{equation}
227: %
228: We substitute this in the representation of the
229: ${}_{3}F_{2}-$function in \pref{P1}
230: %
231: \begin{equation}\label{T1a}
232: \FFF{-n}{\sfrac12}{\sfrac12}{\sfrac12-n}{\sfrac12-n}{-1}=
233: \sum_{k=0}^{n}(-1)^{k}\frac{(-n)_{k}(\frac12)_{k}(\frac12)_{k}}{k!\,(\frac12-n)_{k}(\frac12-n)_{k}}.
234: \end{equation}
235: %
236: This gives after performing the $k-$summation
237: %
238: \begin{equation}\label{T2}
239: f(n)=\frac{n!}{\pi\,(\sfrac12)_n}\int_0^1t^{-\frac12}(1-t)^{n-\frac12}
240: \FG{-n}{\sfrac12}{\sfrac12-n}{\frac{t}{1-t}}\,dt.
241: \end{equation}
242: %
243: We substitute $t=\sin^{2}(\theta/2)$ and obtain
244: %
245: \begin{equation}\label{T3}
246: f(n)=\frac{n!}{\pi\,(\sfrac12)_n}\int_0^{\pi}\cos^{2n}(\theta/2)
247: \FG{-n}{\sfrac12}{\sfrac12-n}{\tan^{2}(\theta/2)}\,d\theta.
248: \end{equation}
249: %
250: We apply a quadratic transformation (see \cite[Eq.~15.3.26]{Abramowitz:1964:HMF}) to obtain
251: %
252: \begin{equation}\label{T4}
253: f(n)=\frac{n!}{\pi\,(\sfrac12)_n}\int_0^{\pi}
254: \FG{-\sfrac12n}{\sfrac12-\sfrac12n}{\sfrac12-n}{\sin^{2}\theta}\,d\theta,
255: \end{equation}
256: %
257: and use the representation of the Legendre polynomial
258: %
259: \begin{equation}\label{T5}
260: P_{n}(x)=\frac{(2n)!}{2^{n}n!\,n!}x^{n}
261: \FG{-\sfrac12n}{\sfrac12-\sfrac12n}{\sfrac12-n}{x^{-2}}.
262: \end{equation}
263: %
264: This follows from \cite[Eq.~(22.3.8)]{Abramowitz:1964:HMF} and gives
265: %
266: \begin{equation}\label{T6}
267: f(n)=\frac{2^{-n}n!\,n!}{\pi\,(\sfrac12)_n\,(\sfrac12)_n}\int_0^\pi
268: \sin^n\theta\ P_n\left(\frac1{\sin\theta}\right)\,d\theta.
269: \end{equation}
270: %
271: Next, consider (see \cite[p.~204]{Temme:1996:SFI})
272: %
273: \begin{equation}\label{T7}
274: P_{n}(z)=\frac1{\pi}\int_{0}^{\pi}
275: \left(z+\sqrt{z^{2}-1}\cos\psi\right)^{n}\,d\psi, \quad
276: n=0,1,2,\ldots,
277: \end{equation}
278: %
279: which gives the double integral
280: %
281: \begin{equation}\label{T8}
282: f(n)=\frac{n!\,n!}{\pi^{2}\,(\sfrac12)_n\,(\sfrac12)_n}\int_0^\pi
283: \int_0^\pi\left(\frac{1+\cos\theta\cos\psi}{2}\right)^n\,d\theta\,d\psi.
284: \end{equation}
285: %
286:
287: \begin{figure}
288: \begin{center}
289: \epsfxsize=6cm \epsfbox{larc.eps}
290: \end{center}
291: \caption{\small The domain of integration of the integral in \pref{T8}
292: and subdomains $A$, $B$, $C$ and $D$. \label{larc1}}
293: \end{figure}
294:
295: %
296: \section{Asymptotic analysis}
297: \label{sec.AA}
298: %
299:
300: The landscape of the integrand in \pref{T8} shows peaks at the boundary points
301: $(0,0)$ and $(\pi,\pi)$, where it assumes the value $1$. Along
302: the interior lines $\theta=\frac12\pi$ and $\psi=\frac12\pi$ the
303: integrand has the value $2^{-n}$. Inside the squares $A$ and $C$, see
304: Figure~\ref{larc1}, the value of the integrand is between $2^{-n}$ and
305: $1$, in the squares $B$ and $D$ it is between $0$ and
306: $2^{-n}$. In addition, the contributions from $A$ and $C$ are the same, and
307: also those from $B$ and $D$ are the same.
308:
309:
310: From an asymptotic point of view it follows that the integral over
311: the full square equals twice the integral over $A$, with an error
312: that is of order $\bigO(2^{-n})$, while the total integral is of order
313: $\bigO(1)$, as $n$ is large. Hence, we concentrate on the integral
314: over $A$, and write for large values of $n$
315: %
316: \begin{equation}\label{A1}
317: f(n)=2\frac{2^{-n}n!\,n!}{\pi^{2}\,(\sfrac12)_n\,(\sfrac12)_n}
318: \left[
319: \int_0^{\frac12\pi}
320: \int_0^{\frac12\pi}(1+\cos\theta\cos\psi)^n\,d\theta\,d\psi+E_{n}\right],
321: \end{equation}
322: %
323: where $E_{n}=\bigO(2^{-n})$. Next, we neglect $E_{n}$ and
324: put $u=\sin(\theta/2), v=\sin(\psi/2)$, and obtain
325: %
326: \begin{equation}\label{A2}
327: f(n)\sim\frac{8\,n!\,n!}{\pi^{2}\,(\sfrac12)_n\,(\sfrac12)_n}
328: \int_0^{\frac12\sqrt{2}}
329: \int_0^{\frac12\sqrt{2}}
330: \left(1-u^{2}-v^{2}+
331: 2u^{2}v^{2}\right)^n\,\frac{du}{\sqrt{1-u^{2}}}\,\frac{dv}{\sqrt{1-v^{2}}}.
332: \end{equation}
333: %
334:
335:
336:
337:
338: For the integrals in \pref{A1} and \pref{A2} asymptotic expansions
339: can be obtained by using Laplace's method for double integrals; see
340: \cite[\S~VIII.10]{Wong:2001:AAI}). In our case a simpler
341: approach is based on neglecting a part of square $A$ by introducing polar coordinates
342: %
343: \begin{equation}\label{A3}
344: u=r\cos t, \quad v=r\sin t,\quad 0\le r\le \sfrac12\sqrt{2}, \quad 0\le t\le
345: \sfrac12\pi.
346: \end{equation}
347: %
348: This gives (again we make an error in the integral that is of order $\bigO(2^{-n})$)
349: %
350: \begin{equation}\label{A4}
351: f(n)\sim\frac{8\,n!\,n!}{\pi^{2}\,(\sfrac12)_n\,(\sfrac12)_n}\int_0^{\pi/2}
352: \int_0^{\frac12\sqrt{2}}\frac{(1-r^{2}+
353: 2r^{4}\cos^{2}t\sin^{2}t)^n r\,dr\,dt}
354: {\sqrt{(1-r^{2}\cos^{2}t)(1-r^{2}\sin^{2}t)}}.
355: \end{equation}
356: %
357: We change $r^{2}$ into $r$, and obtain
358: %
359: \begin{equation}\label{A5}
360: f(n)\sim\frac{4\,n!\,n!}{\pi^{2}\,(\sfrac12)_n\,(\sfrac12)_n}\int_0^{\pi/2}
361: \int_0^{\frac12}\frac{(1-r+
362: 2r^{2}\cos^{2}t\sin^{2}t)^n \,dr\,dt}
363: {\sqrt{(1-r\cos^{2}t)(1-r\sin^{2}t)}}.
364: \end{equation}
365: %
366:
367: First the standard method for obtaining asymptotic expansions of a
368: Laplace-type integral can be used (for the $r-$integral). The second
369: step is done by integrating the coefficients of this expansion with
370: respect to $t$.
371:
372:
373: For the $r-$integral we transform the variable of integration by putting
374: %
375: \begin{equation}\label{A6}
376: w=-\ln\left(1-r+2r^{2}\cos^{2}t\sin^{2}t\right).
377: \end{equation}
378: %
379: This mapping is one-to-one for
380: $r\in[0,\frac12]$, uniformly with respect to $t\in[0,\frac12\pi]$,
381: with corresponding $w-$interval $[0,w_{0}]$, where
382: $w_{0}=w(\frac12)$.
383:
384: We obtain
385: %
386: \begin{equation}\label{A7}
387: f(n)\sim\frac{4\,n!\,n!}{\pi^{2}\,(\sfrac12)_n\,(\sfrac12)_n}\int_0^{\pi/2}
388: \int_0^{w_{0}}e^{-nw}F(w,t)\,dw\,dt,
389: \end{equation}
390: %
391: where
392: %
393: \begin{equation}\label{A8}
394: F(w,t)=\frac{1}{\sqrt{(1-r\cos^{2}t)(1-r\sin^{2}t)}}\,\frac{dr}{dw}.
395: \end{equation}
396: %
397:
398: %
399: \section{Asymptotic expansion}
400: \label{sec.AE}
401: %
402: We obtain the asymptotic expansion of $w-$integral in
403: \pref{A7} by using Watson's lemma (see \cite[\S~I.5]{Wong:2001:AAI}).
404:
405: The function $F(w,t)$ is analytic in a
406: neighborhood of the origin of the $w-$plane. We expand
407: %
408: \begin{equation}\label{E1}
409: F(w,t)=\sum_{k=0}^{\infty} c_{k}(t)w^{k}
410: \end{equation}
411: %
412: and substitute this expansion in \pref{A7}. Interchanging the order
413: of summation and integration, and replacing the interval of the $w-$integrals
414: by $[0,\infty)$ (a standard procedure in asymptotics) we obtain
415: %
416: \begin{equation}\label{E2}
417: f(n)\sim\frac{4\,n!\,n!}{n\pi^{2}\,(\sfrac12)_n\,(\sfrac12)_n}
418: \sum_{k=0}^{\infty}C_{k}\frac{k!}{n^{k}},\quad n\to\infty,
419: \end{equation}
420: %
421: where
422: %
423: \begin{equation}\label{E3}
424: C_{k}=\int_{0}^{\pi/2}c_{k}(t)\,dt,\quad k=0,1,2,\ldots\,.
425: \end{equation}
426: %
427:
428: The coefficients $c_{k}(t)$ can be obtained by the following method.
429: First we need the inverse of the transformation defined in \pref{A6}.
430: That is, we need coefficients $b_{k}$ in the expansion
431: %
432: \begin{equation}\label{E3a}
433: r(w)=\sum_{k=0}^{\infty}b_{k}(t)w^{k}.
434: \end{equation}
435: %
436: We can find $r(w)$ from \pref{A6} as a solution of a quadratic
437: equation, with the condition $r(w)\sim w$ as $w\to 0$, that is,
438: $b_{0}(t)=1$. However, we
439: can also differentiate \pref{A6} with respect to $r$ and substitute
440: the expansion \pref{E3a}, and solve for the coefficients
441: $b_{k}(t)$. When we have these coefficients we can expand $F(w,t)$ of
442: \pref{A8} and find $c_{k}(t)$.
443:
444: The first few coefficients $c_{k}(t)$ are
445: %
446: \renewcommand{\arraystretch}{1.5}
447: \begin{equation}
448: \label{E4}
449: \begin{array}{lll}
450: c_{0}(t) &= & 1,\\
451: c_{1}(t) &= & \frac12( -1+8s^2-8s^4),\\
452: c_{2}(t) &= & \frac18(1-28s^2+220s^4-384s^6+192s^8),\\
453: c_{3}(t) &= & \frac1{48}(-1+92s^2-1628s^4+10752s^6-24576s^8+23040s^{10}-7680s^{12}),\\
454: c_{4}(t) &= &
455: \frac1{384}(1-280s^2+10024s^4-130848s^6+773904s^8-2054400s^{10}+\\
456: & & 2691840s^{12}-1720320s^{14}+430080s^{16}),\\
457: c_{5}(t) &= &
458: \frac1{3840}(-1+848s^2-55328s^4+1259040s^6-13396560s^8+\\
459: & & 73983360s^{10}-215329920s^{12}+349224960s^{14}-\\
460: & & 319549440s^{16}+154828800s^{18}-30965760s^{20})
461: \end{array}
462: \end{equation}
463: \renewcommand{\arraystretch}{1.0}
464: %
465: where $s=\sin^{2}t$. For the corresponding $C_{k}$ we have
466: \begin{equation}
467: \label{E5}
468: C_{0}=\sfrac{1}{2}\pi,\quad
469: C_{1}=0,\quad
470: C_{2}=\sfrac{1}{8}\pi,\quad
471: C_{3}=\sfrac{1}{8}\pi,\quad
472: C_{4}=\sfrac{55}{384}\pi,\quad
473: C_{5}=\sfrac{11}{64}\pi.
474: \end{equation}
475:
476: As a next step we can replace in \pref{E2} the ratios
477: $n!/(\frac12)_n$ by the asymptotic expansion
478: \begin{equation}
479: \label{E6}
480: \frac{n!}{(\frac12)_n}=\sqrt{\pi}\frac{\Gamma(n+1)}{\Gamma(n+\frac12)}\sim
481: \sqrt{\pi n}\sum_{k=0}^{\infty}\frac{\gamma_{k}}{n^{k}},
482: \end{equation}
483: where
484: \begin{equation}
485: \label{E7}
486: \gamma_{0}=1,\quad
487: \gamma_{1}=\sfrac{1}{8},\quad
488: \gamma_{2}=\sfrac{1}{128},\quad
489: \gamma_{3}=-\sfrac{5}{1024},\quad
490: \gamma_{4}=-\sfrac{21}{32768},\quad
491: \gamma_{5}=\sfrac{399}{262144}.
492: \end{equation}
493:
494: This finally gives
495: \begin{equation}
496: \label{E8}
497: f(n)\sim
498: 2\left(1+\frac1{4n}+\frac{17}{32n^{2}}+\frac{207}{128n^{3}}+\frac{14875}{2048n^{4}}+
499: \frac{352375}{8192n^{5}}+\ldots\right).
500: \end{equation}
501:
502:
503: %
504: \section{An alternative method}
505: \label{sec.AM}
506: %
507: The numbers $P_{n}$ were proposed as ``Catalan'' numbers by an associate of
508: Catalan. They appear as coefficients in the series expansion of an
509: elliptic integral of the first kind
510: \begin{equation}
511: \label{M1}
512: K(k)= \int_{0}^{\frac12\pi}\frac{1} {\sqrt{1 -k^2\sin^2t}}\,dt,
513: \end{equation}
514: which is transformed and
515: written as a power series in $k$ (through an intermediate variable);
516: this gives a generating function for the sequence $\{P_{n}\}$.
517: For details we refer to \cite{Jarvis:2004:LRB}.
518:
519: In \cite{Temme:2003:LPC} a generating function for the numbers $P_{n}$
520: is given in terms of the square of a
521: modified Bessel function, and we use this approach to obtain an asymptotic
522: expansion of $f(n)$.
523: See also \cite{Larcombe:2004:ANG} for details on this generating
524: function.
525:
526: We consider numbers $F_{n}$ defined as coefficients in the generating
527: function
528: %
529: \begin{equation}
530: \label{M2}
531: \left[e^{w/2}\,I_0(w/2)\right]^2 = \sum_{n=0}^\infty F_n w^n.
532: \end{equation}
533: %
534: By considering the relation of the Bessel function with the confluent
535: hypergeometric functions (see \cite[Eq.~13.6.3]{Abramowitz:1964:HMF}),
536: \begin{equation}
537: \label{M3}
538: e^{z}I_{\nu}(z)=\frac{(\frac12z)^{\nu}}{\Gamma(\nu+1)}\CHG{\nu+\frac12}{2\nu+1}{2z},
539: \end{equation}
540: %
541: we can write \pref{M2} in the form (see also \cite[Eq.~13.1.27]{Abramowitz:1964:HMF}),
542: %
543: \begin{equation}
544: \label{M4}
545: \left[\CHG{\frac12}{1}{w}\right]^{2}=e^{2w}\left[\CHG{\frac12}{1}{-w}\right]^{2} = \sum_{n=0}^\infty F_n w^n.
546: \end{equation}
547: %
548: This gives the representation for $F_{n}$:
549: %
550: \begin{equation}
551: \label{M5}
552: F_{n}=\sum_{k=0}^{n}\frac{(\frac12)_{k}}{k!\,k!}\,\frac{(\frac12)_{n-k}}{(n-k)!\,(n-k)!}.
553: \end{equation}
554: %
555: By using \pref{P8} it follows that
556: %
557: \begin{equation}
558: \label{M7}
559: F_{n}=\frac{(\frac12)_{n}}{n!\,n!}\sum_{k=0}^{n}(-1)^{k}\frac{(-n)_{k}(-n)_{k}(\frac12)_{k}}{(\frac12-n)_{k}k!\,k!},
560: \end{equation}
561: %
562: or
563: %
564: \begin{equation}
565: \label{M8}
566: F_{n}=\frac{(\frac12)_{n}}{n!\,n!}\FFF{-n}{-n}{\frac12}{1}{\frac12-n}{-1}.
567: \end{equation}
568: %
569: It follows from \pref{P4} that
570: %
571: \begin{equation}
572: \label{M9}
573: f(n)=\frac{n!\,n!\,n!}{2^n\,(\sfrac12)_n\,(\sfrac12)_n} \ F_n.
574: \end{equation}
575: %
576:
577: From \pref{M2} we obtain
578: %
579: \begin{equation}
580: \label{M10}
581: F_{n}=\frac1{2\pi i}\int_{\C}\frac{\left[e^{w/2}\,I_0(w/2)\right]^2}{w^{n+1}}\, dw
582: =\frac1{2\pi i}\int_{\C}\frac{e^{2w}}{w^{n+1}}\,h(w)\, dw,
583: \end{equation}
584: %
585: where
586: %
587: \begin{equation}
588: \label{M11}
589: h(w)=\left[e^{-w/2}\,I_0(w/2)\right]^2=\left[\CHG{\frac12}{1}{-w}\right]^{2},
590: \end{equation}
591: %
592: and the contour $\C$ is a circle around the
593: origin, or any contour that can be obtained from this circle by using
594: Cauchy's theorem. The main contribution comes from the saddle point of
595: $\frac{e^{2w}}{w^{n+1}}$, that is from $w=w_0=n/2$.
596:
597:
598: In the standard saddle point method (see
599: \cite[\S~II.4]{Wong:2001:AAI}) a quadratic transformation is used to
600: bring the main part of the integrand in the form of a Gaussian. We
601: can obtain the same expansion by just expanding the function $h(w)$
602: (which is slowly varying for $w>0$)
603: at the saddle point.
604:
605: First we expand (see \cite[Eq.~13.4.9]{Abramowitz:1964:HMF})
606: %
607: \begin{equation}
608: \label{M12}
609: \CHG{\frac12}{1}{-w}=\sum_{k=0}^{\infty}a_{k}(w-w_{0})^{k},\quad
610: a_{k}=\frac{(-1)^{k}(\frac12)_{k}}{k!\,k!}\CHG{\frac12+k}{1+k}{-w_{0}}
611: \end{equation}
612: %
613: and next
614: %
615: \begin{equation}
616: \label{M13}
617: h(w)=\sum_{k=0}^{\infty}A_{k}(w-w_{0})^{k}.
618: \end{equation}
619: We substitute this expansion in the second integral in \pref{M10}
620: and obtain the convergent expansion
621: %
622: \begin{equation}
623: \label{M14}
624: F_{n}=\sum_{k=0}^{\infty}A_{k}\Phi_{k},\quad \Phi_{k}=\frac1{2\pi
625: i}\int_{\C}\frac{e^{2w}}{w^{n+1}}\,(w-w_{0})^{k}\, dw.
626: \end{equation}
627: %
628: The functions $\Phi_{k}$ can be evaluated by using the recursion
629: formula
630: (which easily follows from integrating by parts)
631: %
632: \begin{equation}
633: \label{M15}
634: \Phi_{k}=-\sfrac12(k-1)(\Phi_{k-1}+w_{0}\Phi_{k-2}),\quad
635: \Phi_{0}=\frac{2^{n}}{n!},\quad \Phi_{1}=0.
636: \end{equation}
637: %
638:
639: An asymptotic expansion can be obtained by using a well-known
640: expansion for
641: $a_{k}$ defined in \pref{M12}. We have (as follows from
642: \cite[Eq.~13.5.1]{Abramowitz:1964:HMF})
643: %
644: \begin{equation}
645: \label{M16}
646: \CHG{a}{c}{-x}\sim
647: x^{-a}\frac{\Gamma(c)}{\Gamma(c-a)}\sum_{m=0}^{\infty}
648: \frac{(a)_{m}(1+a-c)_{m}}{m!\,x^{m}},\quad x\to+\infty,
649: \end{equation}
650: %
651: from which we can obtain expansions for $a_{k}$ and $A_{k}$ for large
652: values of $w_{0}=n/2$. By using these expansions in \pref{M14} we
653: obtain an expansion for $F_{n}$, and finally for $f(n$ by using
654: \pref{M9}. This expansion is the same as the one in \pref{E8}.
655:
656: %
657: \section*{Acknowledgment}
658: I wish to thank Peter Larcombe for suggesting this problem,
659: for encouraging me to investigate the asymptotic properties of the
660: sequence $\{P_{n}\}$, and for introducing me to the literature, in
661: particular to the papers [4] -- [7].
662: %
663:
664:
665: \bibliographystyle{plain}
666:
667: \begin{thebibliography}{10}
668:
669: \bibitem{Abramowitz:1964:HMF}
670: Milton Abramowitz and Irene~A. Stegun.
671: \newblock {\em Handbook of mathematical functions with formulas, graphs, and
672: mathematical tables}, volume~55 of {\em National Bureau of Standards Applied
673: Mathematics Series}.
674: \newblock For sale by the Superintendent of Documents, U.S. Government Printing
675: Office, Washington, D.C., 1964.
676:
677: \bibitem{Bailey:1964:GHS}
678: W.~N. Bailey.
679: \newblock {\em Generalized hypergeometric series}.
680: \newblock Cambridge Tracts in Mathematics and Mathematical Physics, No. 32.
681: Stechert-Hafner, Inc., New York, 1964.
682: \newblock Reprinting of the first edition published in 1935 by Cambridge
683: University Press.
684:
685: \bibitem{Catalan:1887:SNS}
686: E.~Catalan.
687: \newblock Nombres de {S}egner.
688: \newblock {\em Rend. Circ. Mat. Pal.}, 1:190--201, 1887.
689:
690: \bibitem{Clark:2004:AEC}
691: Lane Clark.
692: \newblock An asymptotic expansion for the {C}atalan-{L}arcombe-{F}rench
693: sequence.
694: \newblock {\em J. Integer Seq.}, 7(2):Article 04.2.1, 5 pp. (electronic), 2004.
695:
696: \bibitem{Jarvis:2004:LRB}
697: A.~Frazer Jarvis, Peter~J. Larcombe, and David~R. French.
698: \newblock Linear recurrences between two recent integer sequences.
699: \newblock {\em Congr. Numer.}, 169:79--99, 2004.
700:
701: \bibitem{Larcombe:2006:FPL}
702: Peter~J. Larcombe.
703: \newblock Formal proofs of the limiting behaviour of two finite series using
704: dominated convergence.
705: \newblock Manuscript.
706:
707: \bibitem{Larcombe:2004:ANG}
708: Peter~J. Larcombe and David~R. French.
709: \newblock A new generating function for the {C}atalan-{L}arcombe-{F}rench
710: sequence: proof of a result by {J}ovovic.
711: \newblock {\em Congr. Numer.}, 166:161--172, 2004.
712:
713: \bibitem{Temme:1996:SFI}
714: Nico~M. Temme.
715: \newblock {\em Special functions}.
716: \newblock A Wiley-Interscience Publication. John Wiley \& Sons Inc., New York,
717: 1996.
718: \newblock An introduction to the classical functions of mathematical physics.
719:
720: \bibitem{Temme:2003:LPC}
721: Nico~M. Temme.
722: \newblock Large parameter cases of the {G}auss hypergeometric function.
723: \newblock In {\em Proceedings of the Sixth International Symposium on
724: Orthogonal Polynomials, Special Functions and their Applications (Rome,
725: 2001)}, volume 153, pages 441--462, 2003.
726:
727: \bibitem{Wong:2001:AAI}
728: R.~Wong.
729: \newblock {\em Asymptotic approximations of integrals}, volume~34 of {\em
730: Classics in Applied Mathematics}.
731: \newblock Society for Industrial and Applied Mathematics (SIAM), Philadelphia,
732: PA, 2001.
733: \newblock Corrected reprint of the 1989 original.
734:
735: \end{thebibliography}
736:
737:
738:
739: %\bibliography{larc}
740:
741:
742: \end{document}
743:
744: