1:
2:
3: %%%%%%%%%%%%%%%%%%%%%%%% Ams-Style %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
4: %%%
5: %%% Style and Inputs
6: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
7:
8:
9: \documentclass[10pt]{amsart}
10: \usepackage{amssymb}
11: %\usepackage{CJK,CJKnumb}
12: \usepackage{amsmath,amssymb,amsfonts,amsthm,
13: latexsym, amscd, amsfonts, epsfig}
14: \usepackage{mathrsfs}
15: \usepackage{color}
16: \usepackage{eucal}% caligraphic-euler fonts: \mathcal{ }
17: \usepackage{eufrak}% frak-euler fonts: \mathfrak{ }
18: \usepackage[all]{xypic}
19: \usepackage{xspace}
20:
21: %%%
22: %%%
23: %%%%%%%%%%%%%%%%%%%%%%%%% Pagestyle %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
24: %%%
25: %%%
26:
27: \renewcommand{\baselinestretch}{1.2}% spacing between lines
28:
29: %\hoffset=0truecm
30: %\voffset=0truecm
31: \textwidth=15truecm \textheight=18truecm \baselineskip=0.8truecm
32: \overfullrule=0pt
33: \parskip=0.8\baselineskip
34: \parindent=0truecm
35: \topmargin=0.5truecm \headsep=1.2truecm
36: %\oddsidemargin=0.5in % options for double-side printouts
37: %\evensidemargin=0in
38:
39: %%%
40: %%%
41: %%%%%%%%%%%%%%%%%%%% New Settings %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
42: %%%
43: %%%
44:
45: \theoremstyle{plain}
46: \newtheorem{theorem}{Theorem}
47: \newtheorem{corollary}{Corollary}
48: \newtheorem*{main}{Main~Theorem}
49: \newtheorem{lemma}{Lemma}
50: \newtheorem{proposition}{Proposition}
51: \newtheorem{remark}{Remark}
52: \theoremstyle{definition}
53: \newtheorem{definition}{Definition}
54:
55: \theoremstyle{example}
56: \newtheorem{example}{Example}
57:
58: \theoremstyle{remark}
59:
60: \numberwithin{equation}{section}
61:
62:
63: \begin{document}
64:
65: %%%
66: %%%
67: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
68: %%
69: %%%
70: \title[Asymptotic Enumeration of RNA Structures with Pseudoknots]
71: {Asymptotic Enumeration of RNA Structures with Pseudoknots}
72: \author{Emma Y. Jin and Christian M. Reidys$^{\,\star}$}
73: \address{Center for Combinatorics, LPMC-TJKLC \\
74: Nankai University \\
75: Tianjin 300071\\
76: P.R.~China\\
77: Phone: *86-22-2350-6800\\
78: Fax: *86-22-2350-9272}
79: \email{reidys@nankai.edu.cn}
80: \thanks{}
81: \keywords{Asymptotic enumeration, RNA secondary structure, $k$-noncrossing
82: RNA structure, pseudoknot, generating function, transfer theorem,
83: Hankel contour, singular expansion}
84: \date{June 2007}
85: \begin{abstract}
86: In this paper we present the asymptotic enumeration of RNA structures
87: with pseudoknots. We develop a general framework for the computation
88: of exponential growth rate and the sub exponential factors for
89: $k$-noncrossing RNA structures.
90: Our results are based on the generating function for the number of
91: $k$-noncrossing RNA pseudoknot structures, ${\sf S}_k(n)$, derived in
92: \cite{Reidys:07pseu}, where $k-1$ denotes the maximal size of sets of
93: mutually intersecting bonds.
94: We prove a functional equation for the generating function $\sum_{n\ge
95: 0}{\sf S}_k(n)z^n$ and obtain for $k=2$ and $k=3$ the analytic
96: continuation and singular expansions, respectively.
97: It is implicit in our results that for arbitrary $k$
98: singular expansions exist and via transfer theorems of analytic
99: combinatorics we obtain asymptotic expression for the coefficients.
100: We explicitly derive the
101: asymptotic expressions for $2$- and $3$-noncrossing RNA structures.
102: Our main result is the derivation of the formula ${\sf S}_3(n) \sim
103: \frac{10.4724\cdot 4!}{n(n-1)\dots(n-4)}
104: \left(\frac{5+\sqrt{21}}{2}\right)^n$.
105: \end{abstract}
106: \maketitle
107: {{\small
108: %\tableofcontents
109: }}
110:
111:
112:
113: %%%
114: %%%
115: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
116: %%%
117: %%%
118:
119: \section{Introduction}
120:
121: %%%
122: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
123: %%%
124:
125: RNA molecules are particularly fascinating since they represent
126: both: genotypic legislative via their primary sequence and
127: phenotypic executive via their functionality associated to 2D or
128: 3D-structures, respectively. Accordingly, it is believed that RNA
129: may have been instrumental for early evolution--before Proteins
130: emerged. The primary sequence of an RNA molecule is the sequence of
131: nucleotides {\bf A}, {\bf G}, {\bf U} and {\bf C} together with the
132: Watson-Crick ({\bf A-U}, {\bf G-C}) and ({\bf U-G}) base pairing
133: rules specifying the pairs of nucleotides can potentially form
134: bonds. Single stranded RNA molecules form helical structures whose
135: bonds satisfy the above base pairing rules and which, in many cases,
136: determine their function. For instance, RNA ribozymes are capable of
137: catalytic activity, cleaving other RNA molecules. RNA secondary
138: structure prediction is of polynomial complexity \cite{Waterman:78a}
139: which is result from the fact that in secondary structures no two
140: bonds can cross. Leaving the paradigm of RNA secondary structures,
141: i.e.~studying RNA structures with crossing bonds, the RNA pseudoknot
142: structures, poses challenging problems for computational biology.
143: Prediction algorithms for RNA pseudoknot structures are much harder
144: to derive since there exists no {\it a priori} tree-structure and
145: the subadditivity of local solutions is not guaranteed. RNA
146: pseudoknot structures can be categorized in terms of the maximal
147: size of sets of mutually crossing bonds \cite{Reidys:07pseu}. To be
148: precise a $k$-noncrossing RNA structure has at most $k-1$ mutually
149: crossing bonds and a minimum bond-length of $2$, i.e.~for any $i$,
150: the nucleotides $i$ and $i+1$ cannot form a bond. The asymptotics of
151: $k$-noncrossing RNA structures is of central importance in this
152: context. The key question is how to decompose a $k$-noncrossing RNA
153: structure into a collection of sub-structures (which can easily be
154: computed), and what are the properties of this decomposition. Given
155: such a decomposition we can predict the factors and reassemble the
156: corresponding pseudoknot structure. A first step towards finding
157: such decompositions is to have information about the cardinalities
158: of the respective sets of structures involved.
159: %%%
160: %%%%%%%%%%%%%%%%%%%%%%%% Figures ex1 %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
161: %%%
162: \begin{figure}[ht]
163: \centerline{%
164: \epsfig{file=F11.eps,width=0.8\textwidth}\hskip15pt
165: }
166: \caption{\small RNA secondary structures. Diagram representation
167: (top): the primary sequence, {\bf GAGAGCCUUUGGACCUCA}, is drawn
168: horizontally and its backbone bonds are ignored. All bonds are drawn
169: in the upper halfplane and secondary structures have the property
170: that no two arcs intersect and all arcs have minimum length $2$.
171: Outer planar graph representation (bottom). } \label{F:1}
172: \end{figure}
173: %%%
174: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
175: %%%
176: The asymptotic analysis of $k$-noncrossing RNA structures is based on
177: their generating function, obtained in \cite{Reidys:07pseu}. The particular
178: formulas are however alternating sums, which make even the computation of
179: the exponential growth rate a nontrivial task. In this paper we develop a
180: framework for the asymptotic enumeration of $k$-noncrossing RNA structures.
181: Before we go into this in more detail, let us first provide some background
182: on coarse grained RNA structures and put our results into context.
183:
184: \subsection{RNA secondary structures or the universality of the square root}
185: About three decades ago Waterman {\it et.al.} pioneered the concept
186: of RNA secondary structures
187: \cite{Penner:93c,Waterman:79a,Waterman:78a,Waterman:94a,Waterman:80}.
188: The key property of secondary structures is best understood,
189: considering a structure as a diagram, which is obtained as follows:
190: one draws the primary sequence of nucleotides horizontally and
191: ignores all chemical bonds of its backbone. Then one draws all
192: bonds, i.e.~nucleotide interactions satisfying the Watson-Crick base
193: pairing rules (and {\bf G}-{\bf U} pairs) as arcs in the upper
194: halfplane, effectively identifying structure with the set of all
195: arcs. In this representation, RNA secondary structures have then
196: following property: there exist no two arcs $(i_1,j_1)$,
197: $(i_2,j_2)$, where $i_1<j_1$ and $i_2<j_2$ with the property
198: $i_1<i_2<j_1<j_2$ and all arcs have at least length $2$.
199: Equivalently, there exist no two arcs that cross in the diagram
200: representation of the structure, see Figure~\ref{F:1}.
201: %%%
202: %%%%%%%%%%%%%%%%%%%%%%%% Figures ex1 %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
203: %%%
204: \begin{figure}[ht]
205: \centerline{%
206: \epsfig{file=function.eps,width=0.55\textwidth}\hskip15pt}
207: \caption{\small Universality of the square root. We display the
208: branch-point singularity (here at $\rho_2=\frac{3-\sqrt{5}}{2}$),
209: i.e.~the critical singularity for the asymptotics of RNA secondary structures.
210: {\it All} singularities arising from enumeration of certain classes of
211: secondary structures produces this type, whence the sub exponential factor
212: $n^{-\frac{3}{2}}$.
213: }
214: \label{F:1b}
215: \end{figure}
216: %%%
217: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
218: %%%
219: Basically, all combinatorial properties of secondary structures are derived
220: from Waterman's basic recursion \cite{Waterman:78a}
221: \begin{equation}\label{E:basic}
222: {\sf S}_2(n)= {\sf S}_2(n-1)+\sum_{s=0}^{n-2}
223: {\sf S}_2(n-2-s){\sf S}_2(s) \ ,
224: \end{equation}
225: where ${\sf S}_2(n)$ denotes the number of RNA secondary structures.
226: Eq.~(\ref{E:basic}) is an immediate consequence considering
227: secondary structures as Motzkin paths, i.e.~peak-free paths with
228: {\it up}, {\it down} and {\it horizontal} steps that stay in the
229: upper halfplane, starting at the origin and end on the $x$-axis. The
230: recursion is in particular the key for all asymptotic results since
231: it allows to obtain an implicit function equation for the generating
232: function and subsequent application of Darboux-type theorems
233: \cite{Schuster:98a,Wong:74}. If specific conditions are being
234: imposed, for instance minimum loop-size or stack length, it is
235: straightforward to translate these constraints into restricted
236: Motzkin paths, all of which satisfy some variant of
237: eq.~(\ref{E:basic}). As a result, all asymptotic formulae are of the
238: same type: a square root, that is, the asymptotic behavior is
239: determined by an algebraic branch singularity with the
240: sub exponential factor $n^{-\frac{3}{2}}$. For instance, the number
241: of RNA secondary structures having a minimum hairpin-loop length of
242: $3$ and minimum stack-length $2$ is asymptotically given by ${\sf
243: S}_2(n)\sim 1.4848 \, n^{-\frac{3}{2}} 1.8488^n$
244: \cite{Schuster:98a}. The number of RNA secondary structures having
245: exactly $\ell$ isolated vertices, ${\sf S}_2(n,\ell)$, satisfies the
246: two term recursion $(n-\ell)(n-\ell+2)\, {\sf S}_{2}(n,\ell)\,
247: -\, (n+\ell)(n+\ell-2)\,{\sf S}_{2}(n-2,\ell)=0$
248: \cite{Reidys:07pseu} and Waterman proved in \cite{Waterman:94a} the
249: following beautiful formula
250: \begin{equation}\label{E:Waterman-tree}
251: {\sf S}_2(n,\ell) =
252: \frac{2}{n-\ell}{\frac{n+\ell}{2} \choose \frac{n-\ell}{2} +1}
253: {\frac{n+\ell}{2}-1 \choose \frac{n-\ell}{2}-1}
254: \end{equation}
255: resulting from a bijection between secondary structure and linear trees.
256: In \cite{Waterman:86} it is shown that the prediction of secondary
257: structures can be obtained in polynomial time and yet again
258: eq~(\ref{E:basic}) is central for all folding algorithms
259: \cite{Zuker:79b,Schuster:98a,Waterman:86,Bauer:96,Tacker:94a,McCaskill:90a}.
260:
261: \subsection{Beyond secondary structures}
262: While the concept of secondary structure is of fundamental
263: importance it is well-known that there exist additional types of
264: nucleotide interactions \cite{Science:05a}. These bonds are called
265: pseudoknots \cite{Westhof:92a} and occur in functional RNA (RNAseP
266: \cite{Loria:96a}), ribosomal RNA \cite{Konings:95a} and are
267: conserved in the catalytic core of group I introns.
268: %%%
269: %%%%%%%%%%%%%%%%%%%%%%%% Figures bisec %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
270: %%%
271: \begin{figure}[ht]
272: \centerline{%
273: \epsfig{file=F3.eps,width=0.85\textwidth}
274: }
275: \caption{\small Beyond secondary structures: an RNA bi-secondary structure
276: as the generalization from outer-planar to planar diagrams.
277: We display a secondary RNA structure (top) and a bi-secondary structure
278: (bottom). Reflecting the arcs $(3,8)$ and $(9,12)$ w.r.t. the $x$-axis
279: yields two secondary structures.}
280: \label{F:3}
281: \end{figure}
282: %%%
283: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
284: %%%
285: In plant viral RNAs pseudoknots mimic tRNA structure
286: and in {\it in vitro} RNA evolution \cite{Tuerk:92} experiments have produced
287: families of RNA structures with pseudoknot motifs, when binding HIV-1
288: reverse transcriptase. Important mechanisms like ribosomal frame shifting
289: \cite{Chamorro:91a} also involve pseudoknot interactions.
290: There exist several prediction algorithms for pseudoknot RNA structures
291: \cite{Rivas:99a,Uemura:99a,Akutsu:00a,Lyngso:00a} all of which can identify
292: particular respective pseudoknot motifs.
293: Stadler {\it et al.} \cite{Stadler:99a} suggested a classification of
294: RNA pseudoknot-types based on a notion of inconsistency graphs and
295: computed the upper bound of $4.7613$ for the exponential growth factor
296: of bi-secondary structures. Bi-secondary structures
297: are ``superpositions'' of two secondary structures, i.e.~they can be drawn
298: as a set of non intersecting arcs in the upper and lower half plane,
299: respectively.
300: Figure~\ref{F:3} shows how bi-secondary structures naturally
301: arise when passing from outer-planar to planar diagram representations.
302: The concept of $k$-noncrossing RNA structures generalize both: secondary
303: and bi-secondary structures, respectively.
304: While RNA secondary structures are precisely $2$-noncrossing RNA structures,
305: bi-secondary structures correspond to planar $3$-noncrossing RNA structures.
306: The key advantage of $k$-noncrossing RNA structures is that their defining
307: property is intrinsically local. It can be expected that this facilitates
308: fast folding algorithms. In Figure~\ref{F:4} we contrast all three structural
309: concepts, secondary, bi-secondary and $k$-noncrossing RNA structures.
310: %%%
311: %%%%%%%%%%%%%%%%%%%%%%%% Figures ex1 %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
312: %%%
313: \begin{figure}[ht]
314: \centerline{%
315: \epsfig{file=F4.eps,width=0.7\textwidth}\hskip15pt
316: }
317: \caption{\small $k$-noncrossing RNA structures.
318: (a) secondary structure (with isolated labels $3,7,8,10$),
319: (b) bi-secondary structure, $2,9$ being isolated
320: (c) $3$-noncrossing structure, which is
321: {\it not} a bi-secondary structure In fact, this is {\it the} smallest
322: $3$-noncrossing RNA structure which is not a bi-secondary structure.
323: }
324: \label{F:4}
325: \end{figure}
326: %%%
327: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
328: %%%
329:
330: \subsection{Organization and main results}
331: In Section~\ref{S:pre} provide the necessary background on
332: $k$-noncrossing RNA structures and the generating function
333: $\sum_{n\ge 0}{\sf S}_k(n)z^n$. In Section~\ref{S:exp} we derive the
334: exponential factor for $k$-noncrossing RNA structures, i.e.~we
335: compute the base at which $k$-noncrossing RNA structures
336: asymptotically grow. The exponential factor is {\it the} key result
337: for all complexity considerations arising in the context of
338: prediction algorithms for RNA pseudoknot structures. To make it
339: easily accessible to a broad readership we give an elementary proof
340: based on real analysis and transformations of the generating
341: function. Central to our proof is a functional identity
342: (Lemma~\ref{L:func}) whose true power is revealed only later in
343: Section~\ref{S:sub}, where it is put in the context of analytic
344: functions. Remarkably, Stadler's upper bound for bi-secondary
345: structures coincides with the exact exponential factor obtained via
346: Theorem~\ref{T:asy1} for $3$-noncrossing RNA structures up to
347: $O(10^{-2})$. In Section~\ref{S:sub} we compute the asymptotics for
348: $2$-noncrossing RNA structures and $3$-noncrossing RNA structures,
349: respectively. Since the method via implicit functions used for
350: secondary structures \cite{Schuster:98a} does not work for $k>2$ we
351: develop a new approach which is based on concepts developed by
352: Flajolet {\it et.al.} using singular expansions and transfer
353: theorems
354: \cite{Flajolet:05,Flajolet:99,Flajolet:94,Popken:53,Odlyzko:92}. The
355: basic strategy is as follows: we first obtain an analytic
356: continuation $f(z)$ generalizing the functional equation of
357: Lemma~\ref{L:func} to complex indeterminant $z$. For $k=3$ we obtain
358: an expression involving the Legendre polynomial
359: $P_{\frac{3}{2}}^{-1}(z)$ indicating that the type of singularity is
360: fundamentally different from the branch-point singularity of the
361: square root. In Figure~\ref{F:5} we display the analytic
362: continuation of $\sum_{n\ge 0}{\sf S}_3(n)z^n$ at the dominant
363: singularity, $\rho_3=\frac{5-\sqrt{21}}{2}$ and its singular
364: expansion.
365: %%%%
366: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
367: %%%%
368: \begin{figure}[ht]
369: \centerline{%
370: \epsfig{file=F5.eps,width=1.0\textwidth}\hskip15pt
371: }
372: \caption{\small Toroidal harmonics and its singular expansion. We display
373: the analytic continuation of $\sum_{n\ge 0}{\sf S}_3(n)z^n$, the
374: generating function of $3$-noncrossing RNA structures (left) and its singular
375: expansion (right) at the dominant singularity
376: $\rho_3=\frac{5-\sqrt{21}}{2}$.}
377: \label{F:5}
378: \end{figure}
379: %%%%
380: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
381: %%%%
382: We proceed by proving that $f(z)$ the dominant singularity is indeed unique.
383: The next step is to establish that there exists a singular
384: expansion for $f(z)$, i.e.~there exists a function $h$ such that
385: $f(z)=O(h(z))$ at the dominant singularity (see Section~\ref{S:sub}).
386: Intuitively, this singular expansion approximates $f(z)$ well enough to
387: retrieve precise asymptotic expansions of the coefficients via transfer
388: theorems \cite{Flajolet:05,Gao:92,Handbook:95}.
389: The existence of the singular expansion can be deduced from the particular
390: form of the generating function for $k$-noncrossing RNA structures.
391: Due to Lemma~\ref{L:ana} it suffices to analyze the coefficients
392: $f_3(2n,0)$, which are known via the determinant formula of Bessel
393: functions in eq.~(\ref{E:ww1}). We then proceed using this particular
394: form of $f_3(2n,0)$ to explicitly compute the singular expansion and
395: show in the process how the logarithmic term
396: arises naturally from elementary calculations.
397: It should be remarked that we use the transfer theorems since our
398: generating function is the composition of two analytic functions
399: $f(\vartheta(z))$. We then show that the type of the singualrity
400: of $f(\vartheta(z))$ coincides with the type of singularity of the
401: function $f(z)$. The phenomenon of the persistence of the singularity of
402: the ``outer'' function $f(z)$ is known as the {\it supercritical case}
403: \cite{Flajolet:05}. This will allow us to obtain the asymptotics of
404: the coefficients of the function $f(\vartheta(z))$.
405: One main result of the paper is the formula
406: \begin{eqnarray}
407: {\sf S}_3(n) & \sim & \frac{10.4724\, 4!}{n(n-1)\dots(n-4)}\,
408: \left(\frac{5+\sqrt{21}}{2}\right)^n \ .
409: \end{eqnarray}
410: In order to assess the quality of our formula, let us list
411: the sub exponential factors for $k=2$ and $k=3$, obtained from
412: Theorem~\ref{T:asy2} and
413: Theorem~\ref{T:asy3}:
414: \begin{eqnarray*}
415: {\sf s}_{2}^{}(n) & \sim & 1.1002\,
416: \left[\frac{1}{n^{\frac{3}{2}}}-\frac{7}
417: {8n^{\frac{5}{2}}}-\frac{111}{128n^{\frac{7}{2}}}+
418: \frac{893}{1024n^{\frac{9}{2}}}+{O}(n^{-\frac{11}{2}})\right] \\
419: {\sf s}_{3}^{}(n) & \sim & \frac{10.4724\cdot
420: 4!}{n(n-1)\dots(n-4)}\, \sim 251.3375\left[\frac{1}{
421: n^5}-\frac{35}{4n^6}+\frac{1525}{32 n^7}+{O}(n^{-8})\right] \ .
422: \end{eqnarray*}
423: In the table below we list the sub exponential factors, i.e.~we compare
424: for $k=2,3$ the quantities ${\sf S}_{k}(n)/(\frac{3+\sqrt{5}}{2})^n$ and
425: ${\sf s}_{k}^{}(n)$, respectively. ${\sf S}_2(n)$ and ${\sf S}_3(n)$ are
426: given by the generating function of $k$-noncrossing RNA structures.
427: \begin{center}
428: \begin{tabular}{c|c|c|c|c}
429: \hline
430: \multicolumn{5}{c}{\textbf{The sub exponential factor}}\\
431: \hline
432: $n$ & ${\sf S}_{2}(n)/(\frac{3+\sqrt{5}}{2})^n$ & ${\sf s}_{2}^{}(n)$
433: & ${\sf S}_{3}(n)/(\frac{5+\sqrt{21}}{2})^n$ & ${\sf s}_{3}^{}(n)$\\
434: \hline \small 10 & \small $2.796\times10^{-2}$ & \small $3.124\times10^{-2}$ & \small $5.229\times 10^{-4}$ & \small$1.512\times 10^{-3}$\\
435: \small 20 & \small $1.100\times 10^{-2}$ & \small $1.164\times10^{-2}$ & \small $3.358\times 10^{-5}$ & \small$5.354\times 10^{-5}$\\
436: \small 30 & \small $6.215\times 10^{-3}$ & \small $6.452\times10^{-3}$ & \small $5.776\times 10^{-6}$ & \small$7.874\times 10^{-6}$\\
437: \small 40 & \small $4.114\times 10^{-3}$ & \small $4.229\times10^{-3}$ & \small $1.576\times 10^{-6}$ & \small$1.991\times 10^{-6}$\\
438: \small 50 & \small $2.980\times 10^{-3}$ & \small $3.043\times10^{-3}$ & \small $5.627\times 10^{-7}$ & \small$6.789\times 10^{-7}$\\
439: \small 60 & \small $2.284\times 10^{-3}$ & \small $2.324\times10^{-3}$ & \small $2.397\times 10^{-7}$ & \small$2.804\times 10^{-7}$\\
440: \small 70 & \small $1.822\times 10^{-3}$ & \small $1.849\times10^{-3}$ & \small $1.156\times 10^{-7}$ & \small$1.323\times 10^{-7}$\\
441: \small 80 & \small $1.500\times 10^{-3}$ & \small $1.516\times10^{-3}$ & \small $6.123\times 10^{-8}$ & \small$6.888\times 10^{-8}$\\
442: \small 90 & \small $1.259\times 10^{-3}$ & \small $1.273\times10^{-3}$ & \small $3.483\times 10^{-8}$ & \small$3.868\times 10^{-8}$\\
443: \small 100 & \small $1.078\times 10^{-3}$ & \small $1.088\times10^{-3}$ & \small $2.098\times 10^{-8}$ & \small$2.305\times 10^{-8}$\\
444: \small 1000 & \small $3.484\times 10^{-5}$ & \small $3.475 \times 10^{-5}$ & \small $2.475\times 10^{-13}$ & \small $2.492\times 10^{-13}$ \\
445: \small 10000 & \small $1.104 \times 10^{-6}$ &\small $1.100\times 10^{-6}$ & \small $2.517 \times 10^{-18}$ & \small $2.516 \times 10^{-18}$
446: %\small 20000 & \small $3.904\times 10^{-7}$ &\small $3.904\times 10^{-7}$ & \small $7.874 \times 10^{-20}$ & \small $7.863 \times 10^{-20}$
447: \end{tabular}
448: \end{center}
449:
450:
451: %%%
452: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
453: %%%
454:
455: \section{$k$-noncrossing RNA structures}\label{S:pre}
456:
457: %%%
458: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
459: %%%
460: Suppose we are given the primary RNA sequence
461: $$
462: {\bf A}{\bf A}{\bf C}{\bf C}{\bf A}{\bf U}{\bf G}{\bf U}{\bf G}{\bf G}
463: {\bf U}{\bf A}{\bf C}{\bf U}{\bf U}{\bf G}{\bf A}{\bf U}{\bf G}{\bf G}
464: {\bf C}{\bf G}{\bf A}{\bf C} \ .
465: $$
466: We then identify an RNA structure with the set of all bonds
467: different from the backbone-bonds of its primary sequence, i.e.~the
468: arcs $(i,i+1)$ for $1\le i\le n-1$. Accordingly an RNA structure is
469: a combinatorial graph over the labels of the nucleotides of the
470: primary sequence. These graphs can be represented in several ways.
471: In Figure~\ref{F:6} we represent a particular structure with
472: loop-loop interactions in two ways.
473: %%%%
474: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
475: %%%%
476: \begin{figure}[ht]
477: \centerline{%
478: \epsfig{file=F6.eps,width=0.8\textwidth}\hskip15pt }
479: \caption{\small
480: A $3$-noncrossing RNA structure, as a planar graphs (top) and as a
481: diagram (bottom)} \label{F:6}
482: \end{figure}
483: %%%%
484: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
485: %%%%
486: In the following we will consider structures as diagram
487: representations of digraphs. A digraph $D_n$ is a pair of sets
488: $V_{D_n},E_{D_n}$, where $V_{D_n}=
489: \{1,\dots,n\}$ and $E_{D_n}\subset \{(i,j)\mid 1\le i< j\le n\}$.
490: $V_{D_n}$ and $E_{D_n}$ are called vertex and arc set, respectively.
491: A $k$-noncrossing digraph is a digraph in which all
492: vertices have degree $\le 1$ and which does not contain a $k$-set of
493: arcs that are mutually intersecting, or more formally
494: \begin{eqnarray}
495: \ \not\exists\,
496: (i_{r_1},j_{r_1}),(i_{r_2},j_{r_2}),\dots,(i_{r_k},j_{r_k});\quad & &
497: i_{r_1}<i_{r_2}<\dots<i_{r_k}<j_{r_1}<j_{r_2}<\dots<j_{r_k} \ .
498: \end{eqnarray}
499: We will represent digraphs as a diagrams (Figure~\ref{F:6}) by
500: representing the vertices as integers on a line and connecting any
501: two adjacent vertices by an arc in the upper-half plane. The direction
502: of the arcs is implicit in the linear ordering of the vertices and
503: accordingly omitted.
504: %%%
505: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
506: %%%
507: \begin{definition}\label{D:rna}
508: An $k$-noncrossing RNA structure is a digraph in which all vertices
509: have degree $\le 1$, that does not contain a $k$-set of mutually
510: intersecting arcs and $1$-arcs, i.e.~arcs of the form $(i,i+1)$,
511: respectively. We denote the number of RNA structures by ${\sf
512: S}_k(n)$ and the number of RNA structures with exactly $\ell$
513: isolated vertices and with exactly $h$ arcs by ${\sf S}_k(n,\ell)$
514: and ${\sf S}_k'(n,h)$, respectively. Note that ${\sf S}_k'(n,h)=
515: {\sf S}_k(n,{n-2h})$. We call an RNA structure restricted if
516: and only if it does not contain any $2$-arcs, i.e.~an arc of the
517: form $(i,i+2)$.
518: \end{definition}
519: %%%
520: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
521: Let $f_{k}(n,\ell)$ denote the number of $k$-noncrossing digraphs with
522: $\ell$ isolated points. We have shown in \cite{Reidys:07pseu} that
523: \begin{align}\label{E:ww0}
524: f_{k}(n,\ell)& ={n \choose \ell} f_{k}(n-\ell,0) \\
525: \label{E:ww1}
526: \det[I_{i-j}(2x)-I_{i+j}(2x)]|_{i,j=1}^{k-1} &=
527: \sum_{n\ge 1} f_{k}(n,0)\,\frac{x^{n}}{n!} \\
528: \label{E:ww2}
529: e^{x}\det[I_{i-j}(2x)-I_{i+j}(2x)]|_{i,j=1}^{k-1}
530: &=(\sum_{\ell \ge 0}\frac{x^{\ell}}{\ell!})(\sum_{n \ge
531: 1}f_{k}(n,0)\frac{x^{n}}{n!})=\sum_{n\ge 1}
532: \left\{\sum_{\ell=0}^nf_{k}(n,\ell)\right\}\,\frac{x^{n}}{n!} \ .
533: \end{align}
534: In particular we obtain for
535: $k=2$ and $k=3$
536: \begin{equation}\label{E:2-3}
537: f_2(n,\ell) = \binom{n}{\ell}\,C_{(n-\ell)/2}\quad
538: \text{\rm and}\quad f_{3}(n,\ell)=
539: {n \choose \ell}\left[C_{\frac{n-\ell}{2}+2}C_{\frac{n-\ell}{2}}-
540: C_{\frac{n-\ell}{2}+1}^{2}\right] \ ,
541: \end{equation}
542: where $C_m=\frac{1}{m+1}\binom{2m}{m}$ is the $m$th Catalan number.
543: The derivation of the generating function of $k$-noncrossing RNA structures,
544: given in Theorem~\ref{T:cool1} below uses advanced methods and novel
545: constructions of enumerative combinatorics due to Chen~{\it et.al.}
546: \cite{Chen:07a,Gessel:92a} and Stanley's mapping between matchings and
547: oscillating tableaux i.e.~families of Young diagrams in which any two
548: consecutive shapes differ by exactly one square.
549: The enumeration is obtained using the
550: reflection principle due to Gessel and Zeilberger \cite{Gessel:92a} and
551: Lindstr\"om \cite{Lindstroem:73a} combined with an inclusion-exclusion
552: argument in order to eliminate the arcs of length $1$. In
553: \cite{Reidys:07pseu} generalizations to restricted (i.e. where arcs of
554: the form $(i,i+2)$ are excluded) and circular RNA structures are given.
555: %%%
556: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
557: %%%
558: \begin{theorem}\label{T:cool1}\cite{Reidys:07pseu}
559: Let $k\in\mathbb{N}$, $k\ge 2$, then the
560: number of RNA structures with $\ell$ isolated vertices,
561: ${\sf S}_k(n,\ell)$, is given by
562: \begin{equation}\label{E:da}
563: {\sf S}_k(n,\ell) = \sum_{b=0}^{(n-\ell)/2}
564: (-1)^b\binom{n-b}{b}f_k(n-2b,\ell) \ ,
565: \end{equation}
566: where $f_k(n-2b,\ell)$ is given by the generating function in
567: eq.~{\rm (\ref{E:ww1})}.
568: Furthermore the number of $k$-noncrossing RNA structures, ${\sf S}_k(n)$
569: is
570: \begin{equation}\label{E:sum}
571: {\sf S}_k(n)
572: =\sum_{b=0}^{\lfloor n/2\rfloor}(-1)^{b}{n-b \choose b}
573: \left\{\sum_{\ell=0}^{n-2b}f_{k}(n-2b,\ell)\right\}
574: \end{equation}
575: where $\{\sum_{\ell=0}^{n-2b}f_{k}(n-2b,\ell)\}$ is given by the
576: generating function in eq.~{\rm (\ref{E:ww2})}.
577: \end{theorem}
578: %%%
579: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
580: %%%
581:
582: %%%
583: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
584: %%%
585:
586: \section{The exponential factor}\label{S:exp}
587:
588: %%%
589: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
590: %%%
591: In this section we obtain the exponential growth factor of the
592: coefficients ${\sf S}_k(n)$. Let us begin by considering the
593: generating function $\sum_{n\ge 0}{\sf S}_k(n)z^n$ as a power series
594: over $\mathbb{R}$. Since $\sum_{n\ge 0}{\sf S}_k(n)z^n$ has
595: monotonously increasing coefficients $\lim_{n\to\infty}{\sf
596: S}_k(n)^{\frac{1}{n}}$ exists and determines via Hadamard's formula
597: its radius of convergence. As we already mentioned, due to the
598: inclusion-exclusion form of the terms ${\sf S}_k(n)$, it is not
599: obvious however, how to compute this radius of convergence. Our
600: strategy consists in first showing that ${\sf S}_k(n)$ is closely
601: related to $f_k(2n,0)$ via a functional relation of generating functions.
602: %%%
603: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
604: %%%
605: \begin{lemma}\label{L:func}
606: Let $z$ be an indeterminant over $\mathbb{R}$ and $w\in\mathbb{R}$ a
607: parameter. Let furthermore $\rho_k(w)$ denote the radius of convergence of
608: the power series $\sum_{n\ge 0} [\sum_{h\le n/2} {\sf S}_k(n,h) w^{2h}]
609: z^n$. Then for $\vert z\vert < \rho_k(w)$ holds
610: \begin{equation}\label{E:rr}
611: \sum_{n\ge 0} \sum_{h\le n/2} {\sf S}_k(n,h) w^{2h} z^n =
612: \frac{1}{w^2z^2-z+1}
613: \sum_{n\ge 0} f_k(2n,0) \left(\frac{wz}{w^2z^2-z+1}\right)^{2n} \ .
614: \end{equation}
615: In particular we have for $w=1$,
616: \begin{equation}\label{E:oha}
617: \sum_{n\ge 0} {\sf S}_k(n) z^{n+1} =
618: \sum_{n\ge 0} f_k(2n,0) \left(\frac{z}{z^2-z+1}\right)^{2n+1} \ .
619: \end{equation}
620: \end{lemma}
621: %%%
622: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
623: %%%
624: The proof of Lemma~\ref{L:func} is a bit technical and consists in a series
625: of changes of orders of summations and Laplace transforms. We give the proof
626: in Section~\ref{S:proofs}. In Section~\ref{S:sub} we will employ basic
627: complex
628: analysis and extend eq.~(\ref{E:rr}) to complex $z$. Lemma~\ref{L:func} is
629: the key to prove Theorem~\ref{T:asy1} below, where we obtain the exponential
630: factor for any $k>1$. In its proof we recruit the Theorem of Pringsheim
631: \cite{Titmarsh:39} which asserts that a power series $\sum_{n\ge 0}a_nz^n$
632: with $a_n\ge 0$ has its radius of convergence as dominant (but not
633: necessarily unique) singularity.
634: %%%
635: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
636: %%%
637: \begin{theorem}\label{T:asy1}
638: Let $k$ be a positive integer, $k>1$ and let $r_k$ be the radius of
639: convergence of the power series $\sum_{n\ge 0}f_k(2n,0)z^{2n}$. Then
640: the power series $\sum_{n\ge 0}{\sf S}_k(n)z^n$ has the real valued,
641: dominant singularity at $ \rho_k=
642: \frac{1+\frac{1}{{r_{k}}}}{2}-\sqrt{\left(\frac{1+\frac{1}{{r_{k}}}}
643: {2}\right)^2-1} $ and for the number of $k$-noncrossing RNA
644: structures holds
645: \begin{equation}\label{E:rel}
646: {\sf S}_k(n)\sim \left(\frac{1}{\rho_k}\right)^n \ .
647: \end{equation}
648: \end{theorem}
649: %%%
650: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
651: %%%
652: We will prove later in Theorem~\ref{T:asy2} and Theorem~\ref{T:asy3} that
653: for $k=2$ and $k=3$ the dominant singularities $\rho_2$ and $\rho_3$ are
654: unique, respectively.
655: \begin{proof}
656: Suppose we are given $r_k$, then $r_k\le \frac{1}{2}$ (this follows
657: immediately from $C_n\sim 2^{2n}$ via Stirling's formula) and
658: obviously, $(z-\frac{1}{2})^2+\frac{3}{4}$ has no roots over
659: $\mathbb{R}$. The functional identity of Lemma~\ref{L:func} allows
660: us to derive the radius of convergence of $\sum_{n\ge 0}S_k(n)z^n$.
661: Setting $w=1$ Lemma~\ref{L:func} yields
662: \begin{equation}\label{E:w=1}
663: \sum_{n\ge 0} {\sf S}_k(n) z^n =
664: \frac{1}{(z-\frac{1}{2})^2+\frac{3}{4}}
665: \sum_{n\ge 0} f_k(2n,0) \left(\frac{z}{(z-\frac{1}{2})^2+\frac{3}{4}}\right)
666: ^{2n} \ .
667: \end{equation}
668: $f_k(2n,0)$ is monotone, whence the limit $\lim_{n\to
669: \infty}f_k(2n,0)^{ \frac{1}{2n}}$ exists and applying Hadamard's
670: formula: $\lim_{n\to
671: \infty}f_k(2n,0)^{\frac{1}{2n}}=\frac{1}{{r_{k}}}$. For $z\in
672: \mathbb{R}$, we proceed by computing the roots of
673: $\left|\frac{z}{z^2-z+1}\right|={r_{k}}$ which for $r_k\le
674: \frac{1}{2}$ has the minimal root $\rho_k=
675: \frac{1+\frac{1}{{r_{k}}}}{2}-\left(
676: \left(\frac{1+\frac{1}{{r_{k}}}}{2}\right)^2-1\right)^{\frac{1}{2}}$.
677: We next show that $\rho_k$ is indeed the radius of convergence of
678: $\sum_{n\ge 0} {\sf S}_k(n) z^n$. For this purpose we observe that
679: the map
680: \begin{equation}
681: \vartheta\colon [0, \frac{1}{2}]\longrightarrow [0,\frac{2}{3}],
682: \quad z\mapsto \frac{z}{(z-\frac{1}{2})^2+\frac{3}{4}} , \qquad
683: \text{\rm where} \quad\vartheta(\rho_k)={r_k}
684: \end{equation}
685: is bijective, continuous and strictly increasing.
686: Continuity and strict monotonicity of $\vartheta$ guarantee in view of
687: eq.~(\ref{E:w=1}) that $\rho_k$, is indeed the radius of convergence of
688: the power series $\sum_{n\ge 0} {\sf S}_k(n) z^n$. In order to show that
689: $\rho_k$ is a dominant singularity we consider
690: $\sum_{n\ge 0}{\sf S}_k(n)z^n$ as a power series over $\mathbb{C}$.
691: Since ${\sf S}_k(n)\ge 0$, the theorem of Pringsheim~\cite{Titmarsh:39}
692: guarantees that $\rho_k$ itself is a singularity. By construction
693: $\rho_k$ has minimal absolute value and is accordingly dominant.
694: Since ${\sf S}_k(n)$ is monotone $\lim_{n\to\infty}{\sf S}_k(n)^{\frac{1}
695: {n}}$ exists and we obtain using Hadamard's formula
696: \begin{equation}
697: \lim_{n\to\infty}{\sf S}_k(n)^{\frac{1}{n}}=\frac{1}{\rho_k},\quad
698: \text{\rm or equivalently}\quad {\sf S}_k(n)\sim \left(\frac{1}{\rho_k}
699: \right)^n \, ,
700: \end{equation}
701: from which eq.~(\ref{E:rel}) follows and the proof of the theorem is complete.
702: \end{proof}
703: %%%%
704: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
705: %%%
706: \section{Asymptotics of $3$-noncrossing RNA structures}\label{S:sub}
707: %%%
708: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
709: %%%
710: In this section we provide the asymptotics for RNA secondary and
711: $3$-noncrossing RNA structures. For $k=2$ and $k=3$, i.e.~for RNA
712: secondary and $3$-noncrossing RNA structures, respectively we will
713: explicitly obtain analytic continuations of the power series
714: $\sum_{n\ge 0}{\sf S}_2(n)z^n$ and $\sum_{n\ge 0}{\sf S}_3(n)z^n$,
715: respectively. As a result the dominant singularity relevant for the
716: asymptotics is known and Theorem~\ref{T:asy1} becomes obsolete.
717: However, it is not entirely trivial to derive the analytic
718: continuations for arbitrary crossing numbers $k$. In the context of
719: complexity of prediction algorithms for RNA pseudoknot structures it
720: suffices to obtain the exponential factor which is given via
721: Theorem~\ref{T:asy1}.
722:
723: We begin by revealing the ``true'' power of Lemma~\ref{L:func} in the
724: context of analytic functions.
725: %%%
726: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
727: %%%
728: \begin{lemma}\label{L:ana}
729: Let $k\ge 1$ be an integer, then we have for arbitrary $z\in\mathbb{C}$
730: with the property $\vert z\vert <\rho_k$ the equality
731: \begin{equation}\label{E:rr2}
732: \sum_{n\ge 0} {\sf S}_k(n) z^{n}= \frac{z}{z^2-z+1}
733: \sum_{n\ge 0} f_k(2n,0) \left(\frac{z}{z^2-z+1}\right)^{2n} \ .
734: \end{equation}
735:
736: \end{lemma}
737: %%%
738: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
739: %%%
740: \begin{proof}
741: The power series $\sum_{n\ge 0} {\sf S}_k(n) z^{n}$ and
742: $\sum_{n\ge 0} f_k(2n,0) \left(\frac{z}{z^2-z+1}\right)^{2n}$
743: are analytic in a disc of radius $0<\epsilon<\rho_k$ and
744: according to Lemma~\ref{L:func} coincide on the interval
745: $]-\epsilon,\epsilon [$. Therefore both functions are equal
746: on the sequence $(\frac{1}{n})_{n\in\mathbb{N}}$
747: which converges to $0$ and
748: standard results of complex analysis (zeros of nontrivial
749: analytic functions are isolated) imply that eq.~(\ref{E:rr2})
750: holds for any $z\in\mathbb{C}$ with $\vert z\vert<\rho_k$, whence
751: the lemma.
752: \end{proof}
753:
754: The derivation of the sub exponential factors is based on singular expansions
755: in combination with a transfer theorem, which recruits Hankel contours,
756: see Figure~\ref{F:7}. Let us begin by specifying a suitable domain for our
757: Hankel contours tailored for Theorem~\ref{T:transfer1}.
758: %%%
759: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
760: %%%
761: \begin{definition}\label{D:delta}
762: Given two numbers $\phi,R$, where $R>1$ and $0<\phi<\frac{\pi}{2}$ and
763: $\rho\in\mathbb{R}$ the open domain $\Delta_\rho(\phi,R)$ is defined as
764: \begin{equation}
765: \Delta_\rho(\phi,R)=\{ z\mid \vert z\vert < R, z\neq \rho,\,
766: \vert {\rm Arg}(z-\rho)\vert >\phi\}
767: \end{equation}
768: A domain is a $\Delta_\rho$-domain if it is of the form
769: $\Delta_\rho(\phi,R)$ for some $R$ and $\phi$.
770: A function is $\Delta_\rho$-analytic if it is analytic in some
771: $\Delta_\rho$-domain.
772: \end{definition}
773: %%%
774: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
775: %%%
776: %%%%
777: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
778: %%%%
779: \begin{figure}[ht]
780: \centerline{%
781: \epsfig{file=F7.eps,width=0.5\textwidth}\hskip15pt
782: }
783: \caption{\small $\Delta_1$-domain enclosing a Hankel contour. We
784: assume $z=1$ to be the unique dominant singularity. The coefficients
785: are obtained via Cauchy's integral formula and the integral path is
786: decomposed in $4$ segments. Segment $1$ becomes asymptotically
787: irrelevant since by construction the function involved is bounded on
788: this segment. Relevant are the rectilinear segments $2$ and $4$ and
789: the inner circle $3$. The only contributions to the contour integral
790: are being made here, which shows why the singular expansion allows
791: to approximate the coefficients so well.} \label{F:7}
792: \end{figure}
793: %%%%
794: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
795: %%%%
796: Since the Taylor coefficients have the property
797: \begin{equation}\label{E:scaling}
798: \forall \,\gamma\in\mathbb{C}\setminus 0;\quad
799: [z^n]f(z)=\gamma^n [z^n]f(\frac{z}{\gamma}) \ ,
800: \end{equation}
801: we can, w.l.o.g.~reduce our analysis to the case where $1$ is the dominant
802: singularity. We use the notation
803: \begin{equation}\label{E:genau}
804: \left(f(z)=O\left(g(z)\right) \
805: \text{\rm as $z\rightarrow \rho$}\right)\quad \Longleftrightarrow \quad
806: \left(f(z)/g(z) \ \text{\rm is bounded as $z\rightarrow \rho$}\right) \ .
807: \end{equation}
808: and if we write $f(z)=O(g(z))$ it is implicitly assumed that $z$
809: tends to a (unique) singularity. $[z^n]\,f(z)$ denotes the
810: coefficient of $z^n$ in the power series expansion of $f(z)$ around
811: $0$.
812: %%%
813: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
814: %%%
815: \begin{theorem}\label{T:transfer1}\cite{Flajolet:05}
816: Let $\alpha$ be an arbitrary complex number in $\mathbb{C}\setminus
817: \mathbb{Z}_{\le 0}$ and suppose $f(z)=O((1-z)^{-\alpha})$, then
818: \begin{eqnarray*}
819: [z^n]\, f(z) & \sim & K\ n^{\alpha-1}\left[
820: 1+\frac{\alpha(\alpha-1)}{2n}+\frac{\alpha(\alpha-1)(\alpha-2)(3\alpha-1)}
821: {24 n^2}+\right. \\
822: &&\qquad \quad
823: \left.\frac{\alpha^2(\alpha-1)^2(\alpha-2)(\alpha-3)}{48n^3}+\dots\right]
824: \quad \text{\it for some $K>0$.}
825: \end{eqnarray*}
826: Suppose $r\in\mathbb{Z}_{\ge 0}$, and
827: $f(z)=O((1-z)^{r}\ln^{}(\frac{1}{1-z}))$, then we have
828: \begin{equation}
829: [z^n]f(z)= K\, (-1)^r\frac{r!}{n(n-1)\dots(n-r)} \quad \text{\it for some
830: $K>0$}\ .
831: \end{equation}
832: \end{theorem}
833: %%%
834: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
835: %%%
836: Let us first analyze the case $k=2$, which illustrates the general strategy
837: without the technicality of establishing the existence of a suitable singular
838: expansion. Here the generating function itself can be used directly
839: (i.e.~is its own singular expansion).
840: Our particular proof, given in Section~\ref{S:proofs}, exercises the base
841: strategy used in the proof of Theorem~\ref{T:asy3}. In particular,
842: Theorem~\ref{T:asy2} improves on the quality of approximation providing a
843: sub exponential factor of higher order compared to \cite{Schuster:98a}.
844: %%%
845: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
846: %%%
847: \begin{theorem}\label{T:asy2}
848: The number of RNA secondary i.e.~$2$-noncrossing RNA structures is
849: asymptotically given by
850: \begin{eqnarray} \label{E:konk2}
851: {\sf S}_2(n) \sim
852: \frac{1.1002}{\sqrt{n}}
853: \left(\frac{1}{n+1}-\frac{1}{8n(n+1)}+
854: \frac{1}{128n^{3}}+\frac{5}{1024n^{4}}
855: +{O}(n^{-5})\right) \left(\frac{3+\sqrt{5}}{2}\right)^n \ .
856: \end{eqnarray}
857: \end{theorem}
858: %%%
859: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
860: %%%
861:
862: %%%%%%%%%%%%%%%%%%%%%%%%%%%%% 3-noncrossing %%%%%%%%%%%%%%%%%%%%%%%%%%%%%
863:
864: We next analyze the $3$-noncrossing RNA structures. Here the situation
865: changes dramatically since it has to be shown that a suitable singular
866: expansion exists. We will prove this using the determinant formula
867: arising in the context of the exponential generating function of
868: $f_k(2n,0)$ given in eq.~(\ref{E:ww1}).
869:
870: %%%
871: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
872: %%%
873: \begin{theorem}\label{T:asy3}
874: The number of $3$-noncrossing RNA structures is asymptotically given by
875: \begin{eqnarray*}
876: \label{E:konk3}
877: {\sf S}_3(n) & \sim & \frac{10.4724\cdot 4!}{n(n-1)\dots(n-4)}\,
878: \left(\frac{5+\sqrt{21}}{2}\right)^n \ .\\
879: \end{eqnarray*}
880: \end{theorem}
881: %%%
882: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
883: %%%
884: \begin{proof}
885: {\it Claim $1$.}
886: The dominant singularity $\rho_3$ of the power series
887: $\sum_{n\ge 0} {\sf S}_3(n) z^n$ is unique.\\
888: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
889: In order to prove Claim $1$ we use Lemma~\ref{L:ana}, according to
890: which the analytic function $\Xi_3(z)$ is the analytic continuation
891: of the power series $\sum_{n\ge 0} {\sf S}_3(n) z^n$. We proceed by
892: showing that $\Xi_3(z)$ has exactly $6$ singularities in
893: $\mathbb{C}$, all of which have distinct moduli. The first two
894: singularities are the roots of the quadratic polynomial
895: $P(z)=(z-\frac{1}{2})^2+\frac{3}{4}$, given by
896: $\alpha_1=\frac{1}{2}+i \frac{\sqrt{3}}{2}$ and
897: $\alpha_2=\frac{1}{2}-i\frac{\sqrt{3}}{2}$. Next we observe that the
898: power series $\sum_{n\ge 0} f_k(2n,0) y^{n}$ has the analytic
899: continuation $\Psi(y)$ (obtained by MAPLE sumtools) given by
900: \begin{equation}\label{E:psi}
901: \Psi(y)=
902: \frac{-(1-16y)^{\frac{3}{2}} P_{3/2}^{-1}(-\frac{16y+1}{16y-1})}
903: {16\, {y}^{\frac{5}{2}}} \ ,
904: \end{equation}
905: where $P_{\nu}^{m}(x)$ denotes the Legendre Polynomial of the first kind
906: with the parameters $\nu=\frac{3}{2}$ and $m=-1$.
907: $\Psi(y)$ has one dominant singularity
908: at $y=\frac{1}{16}$, which in view of $\vartheta(z)=(\frac{z}{z^2-z+1})^2$
909: induces exactly $4$ singularities of
910: $
911: \Xi_3(z)=\frac{1}{z^2-z+1}\,
912: \Psi\left(\left(\frac{z}{z^2-z+1}\right)^2\right)
913: $.
914: Indeed, $\Psi(y^2)$ has the two singularities $\mathbb{C}$:
915: $\beta_1=\frac{1}{4}$ and $\beta_2=-\frac{1}{4}$ which produce for
916: $\Xi_3(z)$ the four singularities $\rho_3=\frac{5-\sqrt{21}}{2}$,
917: $\zeta_2=\frac{5+\sqrt{21}}{2}$, $\zeta_3=\frac{-3-\sqrt{5}}{2}$ and
918: $\zeta_4=\frac{-3+\sqrt{5}}{2}$. Therefore all $6$ singularities of
919: $\Xi_3(z)$ have distinct moduli and Claim
920: $1$ follows.\\
921: {\it Claim $2$: the singular expansion}. $\Psi(z)$ is
922: $\Delta_{\frac{1}{16}}(\phi,R)$-analytic and has the singular expansion
923: $(1-16z)^4\ln\left(\frac{1}{1-16z}\right)$.
924: \begin{equation}
925: \forall\, z\in\Delta_{\frac{1}{16}}(\phi,R);\quad
926: \Psi(z)={O}\left((1-16z)^4\ln\left(\frac{1}{1-16z}\right)\right) \ .
927: \end{equation}
928: First $\Delta_{\frac{1}{16}}(\phi,R)$-analyticity of the function
929: $(1-16z)^4\ln\left(\frac{1}{1-16z}\right)$ is obvious.
930: We proceed by proving that $(1-16z)^4\ln\left(\frac{1}{1-16z}\right)$ is
931: the singular expansion.
932: Using the notation of falling factorials
933: $(n-1)_4=(n-1)(n-2)(n-3)(n-4)$ we observe
934: $$
935: f_3(2n,0)=C_{n+2}C_{n}-C_{n+1}^2= \frac{1}{(n-1)_4}
936: \frac{12(n-1)_4(2n+1)}{(n+3)(n+1)^2(n+2)^2}\,
937: \binom{2n}{n}^2 \ .
938: $$
939: With this expression for $f_3(2n,0)$ we arrive at the formal identity
940: \begin{eqnarray*}
941: \sum_{n\ge 5}16^{-n}f_3(2n,0)z^n & = &
942: O(\sum_{n\ge 5}
943: \left[16^{-n}\,\frac{1}{(n-1)_4}
944: \frac{12(n-1)_4(2n+1)}{(n+3)(n+1)^2(n+2)^2}\,
945: \binom{2n}{n}^2-\frac{4!}{(n-1)_4}\frac{1}{\pi}\frac{1}{n}\right]z^n \\
946: & & + \sum_{n\ge 5}\frac{4!}{(n-1)_4}\frac{1}{\pi}\frac{1}{n}z^n) \ ,
947: \end{eqnarray*}
948: where $f(z)=O(g(z))$ denotes that the limit $f(z)/g(z)$ is bounded
949: for $z\rightarrow 1$, eq.~(\ref{E:genau}). It is clear that
950: \begin{eqnarray*}
951: & & \lim_{z\to 1}(\sum_{n\ge 5}\left[16^{-n}\,\frac{1}{(n-1)_4}
952: \frac{12(n-1)_4(2n+1)}{(n+3)(n+1)^2(n+2)^2}\,
953: \binom{2n}{n}^2-\frac{4!}{(n-1)_4}\frac{1}{\pi}\frac{1}{n}\right]z^n) \\
954: &= &
955: \sum_{n\ge 5} \left[16^{-n}\,\frac{1}{(n-1)_4}
956: \frac{12(n-1)_4(2n+1)}{(n+3)(n+1)^2(n+2)^2}\,
957: \binom{2n}{n}^2-\frac{4!}{(n-1)_4}\frac{1}{\pi}\frac{1}{n}\right]
958: <\kappa
959: \end{eqnarray*}
960: for some $\kappa< 0.0784$. Therefore we can conclude
961: \begin{equation}
962: \sum_{n\ge 5}16^{-n}f_3(2n,0)z^n=
963: O(\sum_{n\ge 5}\frac{4!}{(n-1)_4}\frac{1}{\pi}\frac{1}{n}z^n) \ .
964: \end{equation}
965: We proceed by interpreting the power series on the rhs, observing
966: \begin{equation}
967: \forall\, n\ge 5\, ; \qquad
968: [z^n]\left((1-z)^4\,\ln\frac{1}{1-z}\right)=
969: \frac{4!}{(n-1)\dots (n-4)}\frac{1}{n} \, ,
970: \end{equation}
971: whence $\left((1-z)^4\,\ln\frac{1}{1-z}\right)$ is
972: the unique analytic continuation of $\sum_{n\ge 5}\frac{4!}{(n-1)_4}
973: \frac{1}{\pi}\frac{1}{n}z^n$.
974: Using the scaling property of Taylor coefficients
975: $[z^n]f(z)=\gamma^n [z^n]f(\frac{z}{\gamma})$ we obtain
976: \begin{equation}\label{E:isses}
977: \forall\, z\in\Delta_{\frac{1}{16}}(\phi,R);\quad
978: \Psi(z) =O\left((1-16z)^4\ln\left(\frac{1}{1-16z}\right)\right) \ .
979: \end{equation}
980: Therefore we have proved that $(1-16z)^{4}\ln^{}(\frac{1}{1-16z})$ is the
981: singular expansion of $\Psi(z)$ at $z=\frac{1}{16}$, whence Claim $2$.
982: Our last step consists in verifying that the type of the singularity
983: does not change when passing from $\Psi(z)$ to $\Xi_3(z)=\frac{1}{z^2-z+1}
984: \Psi((\frac{z}{z^2-z+1})^2)$. That is, we show that the singular expansion
985: is not affected by substituting $\vartheta(z)=(\frac{z}{z^2-z+1})^2$.
986: \\
987: {\it Claim $3$: the singularity persists}.
988: For $z\in \Delta_{\frac{1}{\rho_3}}(\phi,R)$ we have
989: $\Xi_3(z) ={O}\left((1-\frac{z}{\rho_3})^4\ln(\frac{1}{1-\frac{z}{\rho_3}})
990: \right)$.\\
991: To prove the claim we first observe that Claim $2$ and Lemma~\ref{L:ana}
992: imply
993: \begin{align*}
994: \Xi_3(z)
995: &=O\left(
996: \frac{1}{z^2-z+1}\,\left[\left(1-16(\frac{z}{z^2-z+1})^2\right)^4
997: \ln\frac{1}{\left(1-16(\frac{z}{z^2-z+1})^2\right)}\right]\right) \ .
998: \end{align*}
999: The Taylor expansion of $q(z)=1-16(\frac{z}{z^2-z+1})^2$ at $\rho_3$ is
1000: given by $q(z)=\frac{\sqrt{21}}{5-\sqrt{21}}(\rho_3-z)+{O}(z-\rho_3)^2$ and
1001: setting $\alpha=\frac{\sqrt{21}}{5-\sqrt{21}}$ we compute
1002: \begin{align*}
1003: \frac{1}{z^2-z+1}\,
1004: \left[q(z)^4\ln\frac{1}
1005: {q(z)}\right]
1006: &=
1007: \frac{(\alpha(\rho_3-z)+{O}(z-\rho_3)^2)^4\ln\frac{1}{\alpha(\rho_3-z)+{O}
1008: (z-\rho_3)^2}}{(z-\rho_3)^2+(2\rho_3-1)(z-\rho_3)+\rho_3^2-\rho_3+1}\\
1009: &=
1010: \frac{\left([\alpha+O(z-\rho_3)](\rho_3-z)^4
1011: \ln\frac{1}{[\alpha+O(z-\rho_3)](\rho_3-z)}
1012: \right)}{O(z-\rho_3)+\rho_3^2-\rho_3+1}
1013: \\
1014: &={O}((\rho_3-z)^4\ln\frac{1}{\rho_3-z}) \ ,
1015: \end{align*}
1016: whence Claim $3$. Now we are in the position to employ
1017: Theorem~\ref{T:transfer1}, and obtain for ${\sf S}_3(n)$
1018: \begin{align*}
1019: {\sf S}_{3}(n)&\sim K'\, [z^n]\left((\rho_3-z)^4\ln\frac{1}{\rho_3-z}
1020: \right) \sim K'\, \frac{4!}
1021: {n(n-1)\dots(n-4)}(\frac{1}{\rho_3})^n \ .
1022: \end{align*}
1023: Of course $K'$ can be computed from Theorem~\ref{T:cool1},
1024: explicitly $K'=10.4724$ and the proof of the Theorem
1025: is complete.
1026: \end{proof}
1027:
1028:
1029:
1030: %%%
1031: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1032: %%%
1033:
1034: \section{Proofs}\label{S:proofs}
1035:
1036: %%%
1037: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1038: %%%
1039:
1040:
1041: {\bf Proof of Lemma~\ref{L:func}.}
1042: First we observe that for $z,w\in [-1,1]$ the term $w^2z^2-z+1$ is strictly
1043: positive.
1044: We set
1045: \begin{equation}
1046: F_k(z,w)=\sum_{n\ge 0} \sum_{h\le n/2}{\sf S}_k(n,h)w^{2h}z^n
1047: \end{equation}
1048: and compute
1049: \begin{eqnarray*}
1050: F_k(z,w) & = & \sum_{n\ge 0}\sum_{h\le n/2}\sum_{j=0}^h(-1)^j\binom{n-j}{j}
1051: \binom{n-2j}{2(h-j)} f_k(2(h-j),0)w^{2h}z^n \\
1052: & = & \sum_{n\ge 0}\sum_{j\le n/2}\sum_{h=j}^{n/2}(-1)^j\binom{n-j}{j}
1053: \binom{n-2j}{2(h-j)}f_k(2(h-j),0)w^{2h}z^n \\
1054: & = & \sum_{j\ge 0}\sum_{n\ge
1055: 2j}\sum_{h=j}^{n/2}(-1)^j\binom{n-j}{j}
1056: \binom{n-2j}{2(h-j)}f_k(2(h-j),0)w^{2h}z^n \\
1057: & = & \sum_{j\ge 0}(-1)^j\frac{(wz)^{2j}}{j!}\sum_{n\ge 2j}(n-j)!
1058: \sum_{h=j}^{n/2}\binom{n-2j}{2(h-j)}f_k(2(h-j),0)\frac{w^{2(h-j)}}{(n-2j)!}
1059: z^{n-2j} \ .
1060: \end{eqnarray*}
1061: We shift summation indices $n'=n-2j$ and $h'=h-j$ and derive for the rhs
1062: the following expression
1063: \begin{eqnarray*}
1064: & = & \sum_{j\ge 0}(-1)^j\frac{(wz)^{2j}}{j!}\sum_{n'\ge 0}(n'+j)!
1065: \sum_{h=j}^{n/2}\binom{n'}{2(h-j)}f_k(2(h-j),0)\frac{w^{2(h-j)}}{n'!}
1066: z^{n-2j} \\
1067: & = & \sum_{j\ge 0}(-1)^j\frac{(wz)^{2j}}{j!}\sum_{n'\ge 0}(n'+j)!
1068: \left\{
1069: \sum_{h'=0}^{n/2-j=n'/2}\binom{n'}{2h'}f_k(2h',0)w^{2h'}\right\}
1070: \frac{z^{n'}}{n'!}
1071: \end{eqnarray*}
1072: The idea is now to interpret the term $\sum_{h'=0}^{n'/2}\binom{n'}
1073: {2h'}f_k(2h',0)w^{2h'}\frac{z^n}{n!}$ as a product of the two power series
1074: $e^z$ and $\sum_{n\ge 0}f_k(2n,0)\frac{(wz)^{2n}}{(2n)!}$:
1075: \begin{eqnarray*}
1076: \sum_{\ell\ge 0}\frac{z^\ell}{\ell!}\sum_{n\ge 0}f_k(2n,0)\frac{(wz)^{2n}}
1077: {(2n)!}
1078: & = &
1079: \sum_{n'\ge 0} \sum_{2n+\ell=n'}
1080: \left\{\frac{1}{\ell!}\frac{1}{(2n)!}
1081: f_k(2n,0)w^{2n}\right\}z^{n'}\\
1082: & = & \sum_{n'\ge 0}\left\{\sum_{n=0}^{n'/2}
1083: \binom{n'}{2n}f_k(2n,0)w^{2n}\right\}\frac{z^{n'}}{n'!} \ .
1084: \end{eqnarray*}
1085: We set $\eta_{n'}=\left\{\sum_{h'=0}^{n'/2}\binom{n'}{2h'}f_k(2h',0)
1086: w^{2h'}\right\} $. By assumption we have $\vert z\vert<\rho_k(w)$ and we next
1087: derive, using the Laplace transformation and interchanging integration and
1088: summation
1089: \begin{equation}\label{E:on1}
1090: \sum_{n'\ge 0}(n'+j)!\eta_n
1091: \frac{z^{n'}}{n'!}
1092: = \int_{0}^{\infty}
1093: \sum_{n'\ge 0}\eta_{n'} \frac{(zt)^{n'}}{n'!} t^je^{-t} dt \ .
1094: \end{equation}
1095: Since $\vert z\vert<\rho_k(w)$ the above transformation is valid and using
1096: \begin{equation}
1097: \sum_{n'\ge 0}\left\{\sum_{n=0}^{n'/2}
1098: \binom{n'}{2n}f_k(2n,0)w^{2n}\right\}\frac{z^{n'}}{n'!}=
1099: \sum_{\ell\ge 0}\frac{z^\ell}{\ell!}\sum_{n\ge 0}f_k(2n,0)\frac{(wz)^{2n}}
1100: {(2n)!}
1101: \end{equation}
1102: we accordingly obtain
1103: \begin{eqnarray}
1104: \sum_{n'\ge 0}\eta_{n'} \frac{(zt)^{n'}}{n'!} t^je^{-t} dt & = &
1105: \int_{0}^{\infty}e^{tz}\sum_{n\ge 0}f_k(2n,0)\frac{(wzt)^{2n}}{(2n)!}
1106: t^je^{-t}dt \ .
1107: \end{eqnarray}
1108: The next step is to substitute the term $\sum_{n'\ge 0}(n'+j)!\eta_n
1109: \frac{z^{n'}}{n'!}$ in eq.~(\ref{E:on1}), whence consequently
1110: \begin{eqnarray*}
1111: F_k(z,w) & = & \sum_{j\ge 0}(-1)^j\frac{(wz)^{2j}}{j!}
1112: \int_{0}^{\infty}e^{tz}\sum_{n\ge 0}f_k(2n,0)\frac{(wzt)^{2n}}
1113: {(2n)!}t^je^{-t}dt\\
1114: & = & \int_{0}^{\infty}\sum_{j\ge 0}(-1)^j\frac{(wz)^{2j}}{j!}
1115: e^{tz}\sum_{n\ge 0}f_k(2n,0)\frac{(wzt)^{2n}}
1116: {(2n)!}t^je^{-t}dt \ .
1117: \end{eqnarray*}
1118: The summation over the index $j$ is just an exponential function and we
1119: derive
1120: \begin{eqnarray*}
1121: & = &\int_{0}^{\infty}
1122: e^{-(w^2z^2-z+1)t} \sum_{n\ge 0}f_k(2n,0)\frac{(wzt)^{2n}}
1123: {(2n)!}dt \\
1124: & = &\int_{0}^{\infty}
1125: e^{-(w^2z^2-z+1)t} \sum_{n\ge 0}f_k(2n,0)\frac{1}{(2n)!}
1126: \left(\frac{wz}{w^2z^2-z+1}\right)^{2n} ((w^2z^2-z+1)t)^{2n}dt
1127: \end{eqnarray*}
1128: We proceed by transforming the integral introducing $u=(w^2z^2-z+1)t$,
1129: i.e.~ $dt=(w^2z^2-z+1)^{-1}du$ and accordingly arrive at
1130: \begin{eqnarray*}
1131: F_k(z,w) & = & \sum_{n\ge 0}f_k(2n,0)\frac{1}{(2n)!}
1132: \left(\frac{wz}{w^2z^2-z+1}\right)^{2n}
1133: \int_{0}^{\infty} e^{-(w^2z^2-z+1)t} ((w^2z^2-z+1)t)^{2n}
1134: dt \\
1135: & = & \sum_{n\ge 0}f_k(2n,0)\frac{1}{(2n)!}
1136: \left(\frac{wz}{w^2z^2-z+1}\right)^{2n}
1137: \frac{1}{w^2z^2-z+1} (2n)! \\
1138: & = & \frac{1}{w^2z^2-z+1} \sum_{n\ge 0}f_k(2n,0)
1139: \left(\frac{wz}{w^2z^2-z+1}\right)^{2n} \ ,
1140: \end{eqnarray*}
1141: whence the lemma. $\square$
1142:
1143:
1144: {\bf Proof of Theorem~\ref{T:asy2}.} We shall begin by deriving the
1145: asymptotics of $f_2(2n,0)=C_{n}$. Since $\sum_{n\ge
1146: 0}\binom{2n}{n}z^n=(1-4z)^{-\frac{1}{2}}$, we observe
1147: $$
1148: C_n =\frac{1}{n+1} [z^n]\, (1-4z)^{-\frac{1}{2}}
1149: $$
1150: and according to Theorem~\ref{T:transfer1} we can express $C_n$
1151: asymptotically as
1152: \begin{equation}\label{E:Catalan}
1153: C_{n}\sim \frac{4^{n}}{\sqrt{\pi n}}
1154: \left(\frac{1}{n+1}-\frac{1}{8n(n+1)}+\frac{1}{128n^{3}}+\frac{5}{1024n^{4}}
1155: +{O}(n^{-5})\right) \ .
1156: \end{equation}
1157: The generating function of the Catalan numbers is given by
1158: \begin{equation}\label{E:catalan}
1159: \Psi(y)=\sum_{n\ge 0} C_n y^n =\frac{1-\sqrt{1-4y}}{2y}
1160: \end{equation}
1161: having a branch-point singularity at $\frac{1}{4}$.
1162: Lemma~\ref{L:ana} allows us to express the analytic continuation of
1163: $\sum_{n\ge 0}{\sf S}_2(n)z^n$ via $\Psi$:
1164: \begin{eqnarray}\label{E:oben}
1165: \Xi_2(z)& = &\frac{1}{z^2-z+1}\,
1166: \Psi(\left(\frac{z}{z^2-z+1}\right)^{2}) \\
1167: & = & \frac{1}{z^2-z+1}\
1168: \frac{1-\sqrt{1-4
1169: \left(\frac{z}{z^2-z+1}\right)^{2}}}{2\left(\frac{z}{z^2-z+1}\right)^2}
1170: =\frac{1-\sqrt{1-4
1171: \left(\frac{z}{z^2-z+1}\right)^{2}}}{2\frac{z^2}{z^2-z+1}} \ .
1172: \end{eqnarray}
1173: The explicit form of $\Xi_2(z)$ allows us to conclude that
1174: $\rho_2=\frac{3-\sqrt{5}}{2}$ is the unique dominant singularity. We
1175: denote the map $z\mapsto (\frac{z}{z^2-z+1})^2$ by $\vartheta$ and
1176: compute the first terms of the Taylor series at $z=\rho_2$,
1177: i.e.~where $\vartheta(\rho_2)=\frac{1}{16}$:
1178: \begin{equation}
1179: \vartheta(z) = \frac{1}{4}+\frac{5+3\sqrt{5}}{8}(z-\rho_2) +
1180: (z-\rho_2)^2\,T(z) \ ,
1181: \end{equation}
1182: where $T(z)=\sum_{i\ge 0}c_i(z-\rho_2)^i$, $c_i\in\mathbb{R}$.
1183: Analyzing $\Xi_2(z)$ in an intersection of an $\epsilon$-disc
1184: around $\rho_2$ with $\Delta_{\rho_2}$ produces
1185: \begin{equation}\label{E:what}
1186: \Xi_2(z) = \frac{1-\sqrt{
1187: \left(\frac{5+3\sqrt{5}}{2}\right)(\rho_2-z) -
1188: (z-\rho_2)^2\,T(z)}}{2\frac{z^2}{z^2-z+1}}
1189: \end{equation}
1190: from which we immediately conclude
1191: \begin{equation}
1192: \Xi_2(\rho_2 z)=O(\Psi(4z)) \ .
1193: \end{equation}
1194: Theorem \ref{T:transfer1} and the scaling property of Taylor coefficients
1195: $[z^n]f(z)=\gamma^n [z^n]f(\frac{z}{\gamma})$ imply
1196: \begin{equation}\label{E:coeff}
1197: K [z^n]\, \Xi_2({\rho_2}z) \sim [z^n]\, \Psi({4}z), \quad
1198: \text{\rm for some} \ K>0
1199: \end{equation}
1200: and we accordingly arrive substituting $\alpha=-\frac{1}{2}$ at
1201: \begin{eqnarray*}
1202: \quad [z^n]\, \Xi_2(z) =
1203: \frac{K}{\sqrt{n}}
1204: \left(\frac{1}{n+1}-\frac{1}{8n(n+1)}+\frac{1}{128n^{3}}
1205: +\frac{5}{1024n^{4}}
1206: +{O}(n^{-5})\right)\, \left(\frac{3+\sqrt{5}}{2}\right)^{n},
1207: \end{eqnarray*}
1208: for some $K>0$. Via Theorem~\ref{T:cool1}, the coefficients ${\sf
1209: S}_2(n)$ are explicitly known and we compute $K=1.9572$ from which the
1210: theorem follows. $\square$.
1211:
1212: %%%
1213: %%%
1214: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1215: %%%
1216: {\bf Acknowledgments.}
1217: %%%
1218: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1219: %%%
1220: We are grateful to Prof.~Jason Z.~Gao for helpful comments. Many thanks
1221: to J.Z.M.~Gao
1222: for helping to draw the figures. This work was supported by the 973 Project,
1223: the PCSIRT Project of the Ministry of Education, the Ministry of Science and
1224: Technology, and the National Science Foundation of China.
1225:
1226:
1227: \bibliography{a}
1228: \bibliographystyle{plain}
1229:
1230:
1231: %%%
1232: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1233: %%%
1234:
1235: \end{document}
1236: