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,graphics,

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

32: \textheight=18truecm

33: \baselineskip=0.8truecm

34: \overfullrule=0pt

35: \parskip=0.8\baselineskip

36: \parindent=0truecm

37: \topmargin=0.5truecm

38: \headsep=1.2truecm

39: %\oddsidemargin=0.5in % options for double-side printouts

40: %\evensidemargin=0in

41: 

42: %%%

43: %%%

44: %%%%%%%%%%%%%%%%%%%% New Settings %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

45: %%%

46: %%%

47: 

48: \theoremstyle{plain}

49: \newtheorem{theorem}{Theorem}

50: \newtheorem{corollary}{Corollary}

51: \newtheorem*{main}{Main~Theorem}

52: \newtheorem{lemma}{Lemma}

53: \newtheorem{proposition}{Proposition}

54: \newtheorem{remark}{Remark}

55: \theoremstyle{definition}

56: \newtheorem{definition}{Definition}

57: 

58: \theoremstyle{example}

59: \newtheorem{example}{Example}

60: 

61: \theoremstyle{remark}

62: 

63: \numberwithin{equation}{section}

64: 

65: % references alphabetically

66: % make relation clearer

67: 

68: \begin{document}

69: 

70: %%%

71: %%%

72: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

73: %%

74: %%%

75: \title[Combinatorics of RNA Structures with Pseudoknots]

76:       {Combinatorics of RNA Structures with Pseudoknots}

77: \author{Emma Y. Jin, Jing Qin and Christian M. Reidys$^{\,\star}$}

78: \address{Center for Combinatorics, LPMC \\

79:          Nankai University  \\

80:          Tianjin 300071\\

81:          P.R.~China\\

82:          Phone: *86-22-2350-6800\\

83:          Fax:   *86-22-2350-9272}

84: \email{duck@santafe.edu}

85: \thanks{}

86: \keywords{RNA secondary structure, pseudoknot, enumeration, generating

87: function, reflection principle, walks, Weyl-chamber}

88: \date{April 2007}

89: \begin{abstract}

90: In this paper we derive the generating function of RNA structures with

91: pseudoknots. We enumerate all $k$-noncrossing RNA pseudoknot structures

92: categorized by their maximal sets of mutually intersecting arcs.

93: In addition we enumerate pseudoknot structures over circular RNA.

94: For $3$-noncrossing RNA structures and RNA secondary structures

95: we present a novel $4$-term recursion formula and a $2$-term recursion,

96: respectively.

97: Furthermore we enumerate for arbitrary $k$ all $k$-noncrossing, restricted

98: RNA structures i.e.~$k$-noncrossing RNA structures without $2$-arcs

99: i.e.~arcs of the form $(i,i+2)$, for $1\le i\le n-2$.

100: \end{abstract}

101: \maketitle

102: {{\small

103: %\tableofcontents

104: }}

105: 

106: %%%

107: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

108: %%%

109: %%%

110: 

111: \section{Introduction}

112: 

113: 

114: %%%

115: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

116: %%%

117: 

118: In this paper we study the combinatorics of helical structures of

119: RNA sequences. RNA is described by its primary sequence of nucleotides

120: {\bf A}, {\bf G}, {\bf U} and {\bf C} together with the Watson-Crick

121: ({\bf A-U}, {\bf G-C}) and ({\bf U-G}) base pairing rules specifying

122: which pairs of nucleotides can potentially form bonds. Subject to these

123: single stranded RNA form helical structures.

124: The function of many RNA sequences depends on their structures. Therefore

125: it is important to understand RNA structure in the context of studying the

126: function of biological RNA as well as in the design process of artificial

127: RNA structures. Since RNA is capable of catalytic activity, for instance

128: RNA ribozymes can cleave other RNA molecules, it is believed that RNA may

129: have been instrumental for early evolution, before Proteins emerged.

130: A particularly well-studied sub-class of RNA structures, consisting of

131: planar graphs are the RNA secondary structures.

132: Their combinatorics was pioneered by Waterman {\it et.al.} in a series

133: of seminal papers

134: \cite{Penner:93c,Waterman:79a,Waterman:78a,Waterman:94a,Waterman:80}.

135: RNA secondary structures are coarse grained structures and systematic

136: prediction of the full three dimensional structures, the tertiary

137: structures seems at present time to be out of reach.

138: %%%%%%%%%%%%%%%%%%%%%% Prediction %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

139: It was shown in \cite{Waterman:86} that the prediction of secondary

140: structures can be obtained in polynomial time and their combinatorics,

141: specifically the existence of recursion relations is the key for all

142: folding algorithms \cite{Zuker:79b,Schuster:98}. Over the last two

143: decades a variety of prediction algorithms, based on minimum free

144: energy \cite{Zuker:79b,Waterman:86,Bauer:96}, kinetic folding

145: \cite{Tacker:94a} or the partition function \cite{McCaskill:90a}

146: for RNA secondary structures has been derived.

147: %%%%%%%%%%%%%%%%%%%%%% Why more%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

148: 

149: An increasing number of experimental findings, as well as results from

150: comparative sequence analysis imply that there exist additional types

151: of interactions between RNA nucleotides \cite{Westhof:92a}. These bonds

152: are called pseudoknots and occur in functional RNA like for instance

153: RNAseP \cite{Loria:96a} as well as ribosomal RNA \cite{Konings:95a}.

154: RNA pseudoknots are conserved also in the catalytic core of group I

155: introns. In plant viral RNAs pseudoknots mimic tRNA structure and in

156: {\it in vitro} RNA evolution \cite{Tuerk:92} experiments have produced

157: families of RNA structures with pseudoknot motifs, when binding HIV-1

158: reverse transcriptase. In addition important mechanisms like ribosomal

159: frame shifting \cite{Chamorro:91a} also involve pseudoknot interactions.

160: As a result RNA pseudoknot structures have drawn over the last years a

161: lot of attention \cite{Science:05a}. Several folding algorithms

162: \cite{Rivas:99a,Uemura:99a,Akutsu:00a,Lyngso:00a}

163: have been developed which include certain families of pseudoknots.

164: The prediction problem in general is (although we have not seen formal

165: proof) believed to be NP-hard. In difference to RNA secondary structures

166: a recursive enumeration for pseudoknot RNA is believed to be

167: non-trivial but nevertheless of vital importance for prediction algorithms.

168: Intuitively if bonds can cross it is much harder to enumerate since

169: structural elements can now interact and as a result a structure cannot

170: be straighforwardly decomposed into independent sub-structures.

171: Little is known with respect to the combinatorics of pseudoknot

172: RNA structures.

173: Stadler {\it et al.} \cite{Stadler:99a} suggested a

174: classification of their knot-types based on a notion of inconsistency

175: graphs and provided an upper bound for a certain class of

176: pseudoknots (our 3-noncrossing RNA structures).

177: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

178: 

179: In this paper we introduce a novel approach for the enumeration of

180: RNA structures. Based on new concepts in enumerative combinatorics

181: \cite{Chen:07a,Gessel:92a} we use a method which has the

182: potential to

183: offer insight also into other lattice structure concepts. To be precise

184: Chen~{\it et.al.} have shown in \cite{Chen:07a} that there is a bijection

185: between certain types of matchings and walks inside Weyl-chambers.

186: This bijection is obtained via his construction of oscillating

187: tableaux i.e.~families of Young diagrams in which any two consecutive

188: shapes differ by exactly one square. The corresponding

189: walks can then be enumerated via determinant formulas derived from a

190: reflection principle due to Gessel and Zeilberger

191: \cite{Gessel:92a} and

192: Lindstr\"om \cite{Lindstroem:73a}. The key idea behind the reflection

193: principle is that walks which hit the wall of a Weyl-chamber can be

194: reflected. The original (unreflected) and the reflected walk

195: cancel themselves leaving just the walks that never hit a wall.

196: Crucial for its applicability are restrictive symmetry assumptions

197: since the reflected walk has to be of the same type and, more

198: importantly, the reflection itself can occur at any step.

199: These symmetries are non-existent in walks corresponding to RNA

200: structures. However, our derivation of the generating function of RNA

201: structures is based on these symmetric walks. The key idea is to

202: introduce the asymmetries of RNA structures into the symmetric walks

203: using a certain involution idea. We believe that our particular strategy

204: can be applied for the enumeration of further structure classes.

205: As a result we have tried to keep this paper self contained.

206: 

207: Our main result is the enumeration of all RNA structures. We classify RNA

208: structures by their specific crossing types under the assumption that all

209: base pairs can occur. For arbitrary but fixed $k$ we enumerate all RNA

210: structures with no $k$-set of mutually intersecting bonds.

211: In case of $k=2$, our results

212: reduce to noncrossing structures, i.e.~RNA secondary structures

213: \cite{Waterman:94a} and Waterman's formula for the number of

214: RNA secondary structures with exactly $k$ bonds is derived directly.

215: The case $k=3$ coincides with Stadler's bi-secondary

216: structure \cite{Stadler:99a}. We obtain from the generating

217: function a novel $4$-term recursion formula for RNA structures of length

218: $n$ with no $3$-set of mutually intersecting bonds  and

219: having $\ell$ isolated vertices. We believe that this recursion is

220: the key for developing new prediction algorithms for RNA structures.

221: Also we derive the generating function for circular RNA structures

222: i.e.~for sequences where the bond between $1$ and $n$ is considered part

223: of the primary sequence. Finally we enumerate restricted RNA structures,

224: i.e.~structures in which two interacting nucleotides have at least

225: distance $3$.

226: 

227: The paper is structured as follows. We will begin by introducing several

228: important combinatorial concepts needed for our derivations. Young

229: tableaux, oscillating Young diagrams, RSK algorithm, Weyl-chambers and

230: the reflection principle. We discuss these concepts, illustrate all key

231: ideas and give the corresponding proofs in the Appendix. Our derivation is

232: obtained in three steps. First (Theorem~\ref{T:tableaux}) we show that

233: each structure (represented as a $k$-noncrossing digraph)

234: corresponds uniquely to a walk starting and ending at $(k-1,k-2,

235: \dots,1)$ in $\mathbb{Z}^{k-1}$ and which never touches a wall of the

236: Weyl-chamber $C_0$.

237: Secondly we apply the reflection principle (Theorem~\ref{T:reflect}) in

238: order to count the symmetric walks that remain in the interior of $C_0$.

239: Thirdly (Theorem~\ref{T:cool1}) we incorporate the

240: specific properties of RNA into these symmetric walks and compute the

241: generating function of $k$-noncrossing RNA structures. We show how our

242: results relate to known formulas of RNA secondary structures for which

243: we present a two new term recursion formula. For $3$-noncrossing RNA

244: structures we give a novel $4$-term recursion formula. We finally

245: generalize our strategy (Theorem~\ref{T:cool2}) and enumerate restricted

246: RNA structures.

247: %%%

248: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

249: %%%

250: 

251: \section{From structures to walks and back}

252: 

253: %%%

254: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

255: %%%

256: 

257: Let us begin by illustrating the concept of RNA structures. Suppose we are

258: given the primary sequence

259: $$

260: {\bf A}{\bf A}{\bf C}{\bf C}{\bf A}{\bf U}{\bf G}{\bf U}{\bf G}{\bf G}

261: {\bf U}{\bf A}{\bf C}{\bf U}{\bf U}{\bf G}{\bf A}{\bf U}{\bf G}{\bf G}

262: {\bf C}{\bf G}{\bf A}{\bf C}  \ .

263: $$

264: Structures are combinatorial graphs over the labels of the nucleotides

265: of the primary sequence. These graphs can be represented in several ways.

266: In Figure $1$ we represent a particular structure with loop-loop

267: interactions in two ways:

268: first we display the structure as a planar graph

269: and secondly as a diagram, where the bonds are drawn as arcs in the

270: positive half-plane.

271: %%%%

272: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

273: %%%%

274: \begin{figure}[ht]

275: \centerline{%

276: \epsfig{file=sec1.eps,width=0.9\textwidth}\hskip15pt

277:  }

278: \caption{\small Two representations of RNA structures, planar graphs (top)

279: and diagrams (bottom)}

280: \label{fig:1}

281: \end{figure}

282: 

283: %%%%

284: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

285: %%%%

286: In the following we will consider structures as diagram representations of

287: digraphs.

288: A digraph $D_n$ is a pair of sets $V_{D_n},E_{D_n}$, where $V_{D_n}=

289: \{1,\dots,n\}$ and $E_{D_n}\subset \{(i,j)\mid 1\le i< j\le n\}$.

290: $V_{D_n}$ and $E_{D_n}$ are called vertex and arc set, respectively.

291: A $k$-noncrossing digraph, $G_{k,n}$, is a digraph in which all

292: vertices have degree $\le 1$ and which does not contain a $k$-set of

293: arcs that are mutually intersecting, i.e.

294: \begin{eqnarray}

295: \not\exists\,

296: (i_{r_1},j_{r_1}),(i_{r_2},j_{r_2}),\dots,(i_{r_k},j_{r_k});\quad & &

297: i_{r_1}<i_{r_2}<\dots<i_{r_k}<j_{r_1}<j_{r_2}<\dots<j_{r_k} \ .

298: \end{eqnarray}

299: The set of all $k$-noncrossing digraphs $G_{k,n}$ is denoted by

300: $\mathcal{G}_{n,k}$ and we set ${\sf G}_{n,k}=\vert\mathcal{G}_{n,k}

301: \vert$.

302: The (formal) direction of the edges will have procedural convenience when

303: we map a $k$-noncrossing digraph into an oscillating tableaux

304: (Theorem~\ref{T:tableaux}).

305: We will represent digraphs as a diagrams (Figure~\ref{fig:1}) by

306: representing the vertices as

307: integers on a line and connecting any two adjacent vertices by an arc in

308: the upper-half plane. The direction of the arcs is implicit in the

309: linear ordering of the vertices and accordingly omitted.

310: %%%

311: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

312: %%%

313: \begin{definition}\label{D:rna}

314: An RNA structure (of pseudoknot type $k-2$), ${S}_{k,n}$, is a digraph

315: in which all vertices have degree $\le 1$, that does not contain a

316: $k$-set of mutually intersecting arcs and $1$-arcs, i.e.~arcs of the

317: form $(i,i+1)$, respectively.

318: We denote the number of RNA structures by ${\sf S}_k(n)$ and the number

319: of RNA structures with exactly $\ell$ isolated vertices by ${\sf S}_k(n,

320: \ell)$, respectively.

321: We call an RNA structure restricted iff it does not contain any $2$-arcs,

322: i.e.~an arc of the form $(i,i+2)$.

323: \end{definition}

324: %%%

325: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

326: %%%

327: %%%%%%%%%%%%%%%%%%%%%%%% Figures 1 %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

328: %%%

329: \begin{figure}[ht]

330: \centerline{%

331: \epsfig{file=bi2line.eps,width=0.7\textwidth}\hskip15pt

332:  }

333: \caption{\small RNA structures represented as

334: diagrams, i.e.~arcs over $\{1,\dots,n\}$ in the upper half-plane.

335: $k$-noncrossing digraphs are precisely those which have no $k$-set

336: of mutually intersecting arcs. $2$-noncrossing diagrams without $1$-arcs

337: (top) correspond to secondary structures. $3$-noncrossing diagrams without

338: $1$-arcs (bottom) represent RNA structures with pseudoknots. }

339: \label{fig:2}

340: \end{figure}

341: %%%

342: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

343: %%%

344: We derive the enumeration of RNA structures in three steps.

345: First we establish a bijection from $k$-noncrossing digraphs into a

346: certain class of walks.

347: Secondly we will use the reflection principle in order to count these

348: walks. Thirdly we enumerate all walks subject to specific conditions

349: recruiting a certan involution idea.

350: Let us first discuss two basic concepts needed for our arguments.

351: 

352: {\bf Young tableaux and the RSK algorithm.}

353: A Young diagram (shape) is a collection of squares arranged in

354: left-justified

355: rows with weakly decreasing number of boxes in each row. A Young tableau

356: is a filling of the squares by numbers which is weakly decreasing

357: in each row and strictly decreasing in each column. A tableau is called

358: standard if each entry occurs exactly once. An oscillating tableau is a

359: sequence $\varnothing=\mu^{0}, \mu^{1},\ldots,\mu^{n}=

360: \varnothing$ of standard Young diagrams, such that for

361: $1\le i \le n$, $\mu^{i}$

362: is obtained from $\mu^{i-1}$ by either adding one square or removing

363: one square. For instance the sequence

364: \begin{figure}[ht]

365: \centerline{%

366: \epsfig{file=box.eps,width=0.75\textwidth}\hskip15pt

367:  }

368: \caption{1}

369: \end{figure}

370: 

371: is an oscillating tableaux.

372: 

373: The RSK algorithm is a process of row-inserting elements into a tableau.

374: Suppose we want to insert $k$ into a standard Young tableau $\lambda$.

375: Let $\lambda_{i,j}$ denote the element in the $i$th row and $j$th

376: column of the Young tableau. Let $i$ be the largest integer such that

377: $\lambda_{1,i-1}\le k$. (If $\lambda_{1,1}>k$, then $i=1$.) If

378: $\lambda_{1,i}$ does not exist, then simply add $k$ at the end of the

379: first row. Otherwise, if $\lambda_{1,i}$ exists, then replace

380: $\lambda_{1,i}$ by $k$. Next insert $\lambda_{1,i}$ into the second

381: row following the above procedure and continue until an element is

382: inserted at the end of a row. As a result we obtain a new standard

383: Young tableau with $k$ included. For instance inserting the number

384: sequence $5,2,4,1,6,3$ starting with an empty shape yields the following

385: sequence of standard Young tableaux:

386: \par\medskip

387: 

388: \par\medskip

389: 

390: \par\medskip

391: 

392: \par\medskip

393: \begin{figure}[ht]

394: \centerline{%

395: \epsfig{file=ytl.eps,width=0.75\textwidth}\hskip15pt

396:  }

397: \caption{} \label{fig:4}

398: \end{figure}

399: 

400: {\bf Symmetry groups and Weyl-chambers.} We consider the lattice

401: $\mathbb{Z}^{k-1}$ and walks in $\mathbb{Z}^{k-1}$ having the steps

402: $s$ contained in $\{\pm e_i,0\mid 1\le i\le k-1\}$, where $e_i$ denotes

403: the $i$th unit vector. That is for

404: $a,b\in \mathbb{Z}^{k-1}$ a walk from $a$ to $b$, $\gamma_{a,b}$, of

405: length $n$ is an $n$ tuple $(s_1,\dots,s_n)$ where $s_i\in \{\pm

406: e_i,0 \mid 1\le i\le k-1\}$ such that $b=a+\sum_{h=1}^ns_h$. We set

407: $\gamma_{a,b}(s_r)=a+\sum_{h=1}^rs_h\in \mathbb{Z}^{k-1}$ i.e.~the

408: element at which the walk $(s_1,\dots,s_{r})$ resides at step $r$.

409: The symmetric group $S_{k-1}$ acts on $\mathbb{Z}^{k-1}$ via $\sigma

410: (x_i)_{1\le i\le k-1}= (x_{\sigma^{-1}(i)})_{1\le i\le k-1}$. We set

411: $E_{k-1}=\langle \epsilon_i\mid 1\le i\le k-1\rangle$, where

412: $\epsilon_i(x_1,\dots,

413: x_i,\dots,x_{k-1})=(x_1,\dots,-x_i,\dots,x_{k-1})$. As shown in the

414: Appendix $\{\epsilon_i\sigma\mid \sigma \in S_{k-1},\, \epsilon_i

415: \in E_{k-1}\}$ carries a natural group structure via $(\epsilon_i

416: \sigma) \cdot (\epsilon_j\sigma')=\epsilon_i \sigma \epsilon_j

417: \sigma^{-1}\sigma\sigma'=\epsilon_i\epsilon_{

418: \sigma^{-1}(j)}\sigma\sigma'$. This group, denoted by ${\sf

419: B_{k-1}}$, is generated by $M_{k-1}=\{\epsilon_1,\rho_j\mid 2\le

420: j\le k-1\}$, where $\rho_j=(j,j-1)$, i.e.~$\rho$ transposes the

421: coordinates $x_{j-1}$ and $x_{j}$. By definition ${\sf B}_{k-1}$

422: acts on the set

423: \begin{equation}

424: \Delta_{k-1}=\{\pm e_i \mid 1\le i\le k-1\}\, \cup \,

425: \{e_i \pm e_j \mid 1\le i,j\le k-1\}

426: \end{equation}

427: and we call $\Delta_{k-1}$ the set of roots. We observe that there

428: exists a

429: bijection between $\Delta_{k-1}'=\{e_1, e_j-e_{j-1} \mid 2\le j\le k-1\}$

430: and the set of generators $M_{k-1}$ which maps each root

431: $\alpha\in\Delta_{k-1}'$ into

432: a corresponding reflection (in particular: ${\sf B}_{k-1}$ is generated

433: by reflections)

434: \begin{equation}\label{E:Co}

435: \{ e_1, e_j-e_{j-1}\mid 2\le j\le k-1 \} \longrightarrow

436: \{ \epsilon_1,\rho_j\mid 2\le j\le k-1\} , \quad \alpha \mapsto

437: \left(\beta_\alpha= x \mapsto x-2 \frac{\langle \alpha,x\rangle }

438:           {\langle\alpha,\alpha\rangle}\right)

439: \end{equation}

440: where $\langle x,x'\rangle$ denotes the standard scalar product in

441: $\mathbb{Z}^{k-1}$. It is clear that $\Delta_{k-1}'$ is a basis of

442: $\mathbb{Z}^{k-1}$. We refer to the sub spaces $\langle e_i\rangle$

443: for $1\le i\le k-1$ and $\langle e_j-e_{j-1}\rangle$ for $2\le j\le

444: k-1$ as walls. A ${\sf B}_{k-1}$-chamber is defined as the set of

445: $x\in \mathbb{Z}^{k-1}$ with the property that $\langle

446: \alpha,x\rangle \ge 0$ for all $\alpha \in \Delta_{k-1}$.

447: We denote the Weyl chamber

448: \begin{equation}\label{E:C_0}

449: C_0=\{x\in\mathbb{Z}^{k-1}\mid 0<x_{k-1}<x_{k-2}<\dots<x_1\} \ .

450: \end{equation}

451: For RNA secondary structures we have $k-1=1$, and

452: ${\sf B}_1=E_1=\{\epsilon_1, 1\}$ and $\Delta'=\{e_1\}$.

453: For $3$-noncrossing RNA we have $k-1=2$ and

454: ${\sf B}_2=E_2 \rtimes_\varphi S_2\cong D_4$ (where $\varphi:S_2\to

455: {\sf Aut}(E_2)$) is the dihedral group of order $8$.

456: 

457: The following theorem is the first step for the enumeration of RNA

458: structures. It will allow to interpret a certain class of digraphs

459: as walks in $\mathbb{Z}^{k-1}$ which remain in the interior of the Weyl

460: chamber $C_0$. The result is due to Chen~{\it et al} \cite{Chen:07a},

461: where it is formulated for matchings. The original bijection between

462: oscillating tableaux and matchings is due to Stanley and was

463: generalized Sundaram \cite{Sundaram:90a}.

464: We give a proof of Theorem~\ref{T:tableaux} in the Appendix.

465: %%%

466: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

467: %%%

468: \begin{theorem}\cite{Chen:07a}\label{T:tableaux}

469: There exists a bijection between $k$-noncrossing digraphs and walks

470: of length $n$ in $\mathbb{Z}^{k-1}$ which start and end at $a=(k-1,k-2,

471: \dots,1)$ having steps $0,\pm e_i$, $1\le i\le k-1$ such that $0<

472: x_{k-1}<\dots < x_1$ at any step.

473: I.e.~we have a bijection

474: \begin{equation}\label{E:bij3}

475: \mathcal{G}_{n,k} \longrightarrow

476: \{\gamma_{a,a}\mid \gamma_{a,a} \,

477:                     \text{\it remains inside the Weyl-chamber $C_0$}\} \ ,

478: \end{equation}

479: where $\mathcal{G}_{n,k}$ denotes the set of $k$-noncrossing digraphs of

480: length $n$.

481: \end{theorem}

482: %%%

483: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

484: %%%

485: \begin{figure}[ht]

486: \centerline{%

487: \epsfig{file=bioneline.eps,width=0.75\textwidth}\hskip15pt

488:  }

489: \caption{\small A $5$-noncrossing digraph. In the text we show how to

490: derive from this digraph an oscillating tableau and subsequently the

491: corresponding walk $\gamma_{a,a}$ in $\mathbb{Z}^4$.}

492: \label{fig:3}

493: \end{figure}

494: \begin{figure}[ht]

495: \centerline{%

496: \epsfig{file=2tl.eps,width=0.75\textwidth}\hskip15pt

497:  }

498: \caption{1} \label{fig:5}

499: \end{figure}

500: 

501: The $5$-noncrossing digraph in Figure~\ref{fig:3} corresponds to a

502: oscillating tableau as follows:

503: from right to left start at vertex $17$, which is a terminus. The

504: corresonding origin is $7$ which is inserted via the RSK algorithm

505: into the empty shape. Next insert the origin corresponding to $16$ and

506: $15$, respectively. At $14$ and $13$ nothing happens since they are

507: isolated vertices. At $12$ origin $3$ is inserted and $11$ is isolated.

508: $10$ is an origin of an arc and accordingly removed from the tableaux.

509: It is now clear how to proceed. The walk $\gamma_{a,a}$ is obtained

510: from the tableau as follows: its $x_i$-coordinate is the number of

511: squares in the $i$-th column, i.e.~$\gamma_{a,a}$ is given by

512: {\small

513: \begin{align*}

514: &(4,3,2,1),(5,3,2,1),(6,3,2,1),(6,4,2,1),(6,4,2,1),(6,4,2,1),(6,4,3,1),

515: (6,4,3,1),(5,4,3,1),\\

516: &(5,4,3,2),(6,4,3,2),(6,4,3,1),(6,4,3,1),(6,4,2,1),(6,4,2,1),(6,3,2,1),

517: (5,3,2,1),(4,3,2,1) \ .

518: \end{align*}}

519: We show in the appendix in detail why this is a bijection.

520: 

521: %%%

522: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

523: %%%

524: We next discuss the reflection principle. The key idea is to count

525: walks that remain in the interior of a Weyl chamber by counting all

526: walks. Then one utilizes the fact that all walks that touch a wall

527: at some step can be paired and eventually cancel themselves in the

528: enumeration. The particular way to obtain this pairing is by reflecting

529: the walk at the corresponding wall. The following observation is essential

530: for the reflection principle, formulated in Theorem~\ref{T:reflect} below.

531: %%%

532: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

533: %%%

534: \begin{lemma}\label{L:touch}

535: Let  $\Delta_{k-1}'=\{ e_1, e_j-e_{j-1}\mid 2\le j\le k-1 \}$. Then

536: every walk starting at some lattice point in the interior of $C$

537: having steps $\pm e_i,0$ that crosses from inside ${C}$ into outside

538: ${C}$ touches a subspace $\langle e_i- e_{i-1}\mid 2\le i\le

539: k-1\rangle$ or $\langle e_i\mid 1\le i\le k-1\rangle$.

540: \end{lemma}

541: %%%

542: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

543: %%%

544: Let $\Gamma_n(a,b)$ be the number of walks $\gamma_{a,b}$. For

545: $a,b\in C_0$ (eq.~(\ref{E:C_0}))

546: let $\Gamma_n^+(a,b)$ denote the number of walks $\gamma_{a,b}$ that never

547: touch a wall, i.e.~remain in the interior of $C_0$. Finally for

548: $a,b\in \mathbb{Z}^{k-1}$, let $\Gamma_n^-(a,b)$ denote the number of walks

549: $\gamma_{a,b}=(s_1,\dots,s_n)$ that hit a wall at some step $s_r$.

550: $\ell(\beta)$ denotes the length of $\beta\in{\sf B}_{k-1}$.

551: For $a=b=(k-1,\dots,1)$ we have according to Theorem~\ref{T:tableaux}

552: \begin{equation}

553: \Gamma_n^+(a,a) = {\sf G}_{n,k}\ ,

554: \end{equation}

555: where ${\sf G}_{n,k}=\vert \mathcal{G}_{n,k}\vert$.

556: %%%

557: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

558: %%%

559: \begin{theorem}{\bf (Reflection-Principle)}\cite{Gessel:92a}

560: \label{T:reflect}

561: Suppose $a,b\in C_0$, then we have

562: \begin{equation}

563: \Gamma_n^+(a,b)=

564:            \sum_{\beta \in {\sf B}_{k-1}}(-1)^{\ell(\beta)} \,

565:                   \Gamma_n(\beta(a),b) \ .

566: \end{equation}

567: \end{theorem}

568: %%%

569: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

570: %%%

571: Theorem \ref{T:reflect} allows us to compute the exponential

572: generating function for $\Gamma_{n}^{+}(a,b)$, which is the number of

573: walks from $a$ to $b$, that remain in the interior of $C_{0}$

574: \cite{Grabiner:93a}.

575: %%%

576: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

577: %%%

578: \begin{lemma}\label{L:walks}\cite{Grabiner:93a}

579: Let $I_{r}(2x)=\sum_{j \ge 0}x^{2r+j}/{j!(r+j)!}$ be the

580: hyperbolic Bessel function of the first kind of order $r$. Then the

581: generating functions for the numbers of $k$-noncrossing digraphs of

582: length $n$ and for $k$-noncrossing digraphs of length $n$ without

583: isolated points, $\Gamma_n^+(a,b)$ and ${\Gamma'}_n^+(a,b)$ are given by

584: \begin{eqnarray}

585: \sum_{n \ge 0}\Gamma_{n}^{+}(a,b)\frac{x^{n}}{n!}& =& e^{x}\det

586: [I_{a_{i}-b_{j}}(2x)-I_{a_{i}+b_{j}}(2x)]|_{i,j=1}^{k-1} \\

587: \sum_{n \ge 0}{\Gamma'}_{n}^{+}(a,b)\frac{x^{n}}{n!}& =& \det

588: [I_{a_{i}-b_{j}}(2x)-I_{a_{i}+b_{j}}(2x)]|_{i,j=1}^{k-1} \ .

589: \end{eqnarray}

590: \end{lemma}

591: %%%

592: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

593: %%%

594: Now we can combine our results in order to enumerate $k$-noncrossing

595: digraphs using the bijection between digraphs and walks.

596: Theorem~\ref{T:tableaux} implies that the number of $k$-noncrossing

597: digraphs is equal to $\Gamma^+_{k}(a,a)$, the number of lattice walks

598: in $\mathbb{Z}^{k-1}$ of length $n$ that remain in the interior of

599: $C_{0}$ from $(k-1,\ldots 1)$ to itself with steps $0,\pm e_{i}$, $1

600: \le i \le k-1$. According to Lemma~\ref{L:walks} the generating

601: functions for walks with steps $e_i=\pm 1,0$ and $e_i=\pm 1$,

602: where $1\le i\le k-1$ are given by

603: \begin{equation}\label{E:gen}

604: e^{x}\det_{}[I_{i-j}(2x)-I_{i+j}(2x)]|_{i,j=1}^{k-1} \quad

605: \text{\rm and}\quad \det[I_{i-j}(2x)-I_{i+j}(2x)]|_{i,j=1}^{k-1} \ ,

606: \end{equation}

607: respectively.

608: Let $f_{k}(n,\ell)$ denote the number of $k$-noncrossing digraphs with

609: $\ell$ isolated points. Then

610: \begin{align}\label{E:ww0}

611: f_{k}(n,\ell)& ={n \choose \ell} f_{k}(n-\ell,0) \\

612: \label{E:ww1}

613: \det[I_{i-j}(2x)-I_{i+j}(2x)]|_{i,j=1}^{k-1} &=

614: \sum_{n\ge 1} f_{k}(n,0)\cdot\frac{x^{n}}{n!} \\

615: \label{E:ww2}

616: e^{x}\det[I_{i-j}(2x)-I_{i+j}(2x)]|_{i,j=1}^{k-1}

617: &=(\sum_{\ell \ge 0}\frac{x^{\ell}}{\ell!})(\sum_{n \ge

618: 1}f_{k}(n,0)\frac{x^{n}}{n!})=\sum_{n\ge 1}

619: \left\{\sum_{\ell=0}^nf_{k}(n,\ell)\right\}\cdot\frac{x^{n}}{n!} \ .

620: \end{align}

621: In particular we obtain for

622: $k=2$ and $k=3$

623: \begin{equation}\label{E:2-3}

624: f_2(n,\ell)  =  \binom{n}{\ell}\,C_{(n-\ell)/2}\quad

625: \text{\rm and}\quad  f_{3}(n,\ell)=

626: {n \choose \ell}\left[C_{\frac{n-\ell}{2}+2}C_{\frac{n-\ell}{2}}-

627:       C_{\frac{n-\ell}{2}+1}^{2}\right] \ ,

628: \end{equation}

629: where $C_m$ denotes the $m$th Catalan number.

630: %%%

631: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

632: %%%

633: \section{RNA structures}\label{S:RNA}

634: %%%

635: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

636: %%%

637: 

638: In this section we derive the generating function for RNA structures.

639: The successful strategy consists in counting the ``wrong'' object

640: ``multiple'' times.

641: To be precise we will enumerate all $k$-noncrossing digraphs with $j$

642: $1$-arcs by relating them to a {\it family} whose cardinality

643: we can easily compute.

644: We denote the number of RNA structures with exactly $\ell$ isolated

645: vertices by ${\sf S}_k(n,\ell)$. Suppose $k\ge 2$

646: and let $\mathcal{G}^{}_{n,k}(\ell,j)$ be the set of all $k$-noncrossing

647: digraphs having exactly $\ell$ isolated points and exactly $j$ $1$-arcs,

648: where a $1$-arc is an arc of the form $(i,i+1)$, $1\le i\le n-1$. Setting

649: $

650: {\sf G}_{k}(n,\ell,j)=\vert \mathcal{G}_{n,k}(\ell,j)\vert

651: $,

652: we have in particular ${\sf S}_k(n,\ell)={\sf G}_{k}(n,\ell,0)$.

653: %%%

654: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

655: %%%

656: \begin{theorem}\label{T:cool1}

657: Let $k\in\mathbb{N}$, $k\ge 2$, let $C_m$ denote the $m$-th Catalan number

658: and $f_k(n,\ell)$ be the number of $k$-noncrossing digraphs over $n$

659: vertices with exactly $\ell$ isolated vertices. Then the

660: number of RNA structures with $\ell$ isolated vertices,

661: ${\sf S}_k(n,\ell)$, is given by

662: \begin{equation}\label{E:da}

663: {\sf S}_k(n,\ell) = \sum_{b=0}^{(n-\ell)/2}

664:                    (-1)^b\binom{n-b}{b}f_k(n-2b,\ell)  \  ,

665: \end{equation}

666: where $f_k(n-2b,\ell)$ is given by the generating function in

667: eq.~{\rm (\ref{E:ww1})}.

668: Furthermore the number of $k$-noncrossing RNA structures, ${\sf S}_k(n)$

669: is

670: \begin{equation}\label{E:sum}

671: {\sf S}_k(n)

672: =\sum_{b=0}^{\lfloor n/2\rfloor}(-1)^{b}{n-b \choose b}

673: \left\{\sum_{\ell=0}^{n-2b}f_{k}(n-2b,\ell)\right\}

674: \end{equation}

675: where $\{\sum_{\ell=0}^{n-2b}f_{k}(n-2b,\ell)\}$ is given by the

676: generating function in eq.~{\rm (\ref{E:ww2})}.

677: \end{theorem}

678: %%%

679: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

680: %%%

681: \begin{proof}

682: We first prove

683: \begin{equation}\label{E:form1}

684: \sum_{j\ge b}\binom{j}{b}\, {\sf G}_k(n,\ell,j)=

685: \binom{n-b}{b}

686: \, f_k(n-2b,\ell) \ .

687: \end{equation}

688: For this purpose we construct a family $\mathcal{F}$ of

689: $\mathcal{G}_{n,k}$-digraphs, having exactly $\ell$ isolated

690: points and having at least $b$ $1$-arcs as follows:

691: select {\sf (a)} $b$ $1$-arcs, and {\sf (b)} an arbitrary

692: $k$-noncrossing digraph with exactly $\ell$ isolated points

693: over the remaining $n-2b$ vertices. Let $\mathcal{F}$ be the

694: resulting family of digraphs. \\

695: {\it Claim $1$.} Each element $\theta\in \mathcal{F}$ is contained in

696: $\mathcal{G}_{n,k}(\ell,j)$ for some $j\ge b$.\\

697: To prove this we observe that a $1$-arc cannot cross any other arc,

698: i.e.~cannot be contained in a set of mutually crossing arcs. As a

699: result for $k\ge 2$ our construction generates digraphs that are

700: $k$-noncrossing. Clearly $\theta$ has exactly $\ell$ isolated

701: vertices and in step {\sf (b)} we potentially derive additional

702: $1$-arcs, whence $j\ge b$.\\

703: {\it Claim $2$.}

704: \begin{equation}\label{E:fam}

705: \vert \mathcal{F}\vert =\binom{n-b}{b}f_k(n-2b,\ell)\ .

706: \end{equation}

707: Let $\lambda(n,b)$ denote the number of ways to select $b$ $1$-arcs

708: over $\{1,\dots,n\}$. We observe that $\lambda(n,b)=\binom{n-b}{b}$.

709: Identifying the two incident vertices of an $1$-arc we conclude that we

710: can choose the $b$ $1$-arcs in $\binom{n-b}{b}$ ways.

711: Obviously, $\ell$ isolated vertices can be obtained in $\binom{n-2b}

712: {\ell}$ different ways and it remains to select an arbitrary

713: $k$-noncrossing digraph with exactly $\ell$ isolated points over $n-2b$

714: vertices. The number of those is given by $f(n-2b,\ell)$ which we can

715: compute via Lemma~\ref{L:walks}, whence eq.~(\ref{E:fam}) and Claim $2$ is

716: proved.\\

717: In view of the fact that any of the $k$-noncrossing digraphs can introduce

718: additional $1$-arcs we set

719: $$

720: \mathcal{F}(j)=\{\theta\in\mathcal{F} \mid \theta\ \text{\rm has

721:                      exactly $j$ $1$-arcs} \} \ .

722: $$

723: Obviously, $\mathcal{F}=\dot\bigcup_{j \ge b}\mathcal{F}(j)$.

724: Suppose $\theta \in \mathcal{F}(j)$. According to Claim $1$, $\theta\in

725: \mathcal{G}_{n,k}(\ell,j)$ and furthermore $\theta$ occurs with

726: multiplicity $\binom{j}{b}$ in $\mathcal{F}$ since by construction any

727: $b$-element subset of the $j$ $1$-arcs is counted respectively in

728: $\mathcal{F}$. Therefore we have

729: \begin{equation} \label{E:well}

730: \vert \mathcal{F}(j)\vert = \binom{j}{b}{\sf G}_k(n,\ell,j)

731: \end{equation}

732: and

733: \begin{eqnarray*}

734: \sum_{j \ge b}\binom{j}{b}{\sf G}_k(n,\ell,j)  =

735: \sum_{j \ge b} \vert \mathcal{F}(j)\vert =

736: \binom{n-b}{b} f_k(n-2b,\ell) \ ,

737: \end{eqnarray*}

738: whence eq.~(\ref{E:form1}). We next set

739: $F_k(x)=\sum_{j\ge 0}{\sf G}_k(n,\ell,j) \, x^j$.

740: Taking the $b$-th derivative and let $x=1$ we obtain

741: \begin{eqnarray}\label{E:u}

742: \frac{1}{b!} F_k^{(b)}(1) =

743: \sum_{j\ge b}\binom{j}{b}{\sf G}_k(n,\ell,j)1^{j-b}

744:  \ .

745: \end{eqnarray}

746: Claim~$2$ provides an interpretation of

747: the r.h.s.~of eq.~(\ref{E:u})

748: \begin{equation}

749: \sum_{j\ge b}\binom{j}{b}{\sf G}_k(n,\ell,j)\, 1^{j-b}=

750: \binom{n-b}{b}f_k(n-2b,\ell) \ .

751: \end{equation}

752: In order to connect $F_k(x)$ and $\frac{1}{b!} F^{(b)}(1)$ we consider the

753: Taylor expansion of $F_k(x)$ at $x=1$ and compute

754: \begin{eqnarray*}

755: F_k(x)  =  \sum_{b\ge 0} \frac{1}{b!} \, F^{(b)}(1) (x-1)^b

756:         =  \sum_{b=0}^{(n-\ell)/2}\binom{n-b}{b} f_k(n-2b,\ell)

757:               (x-1)^b \ .\\

758: \end{eqnarray*}

759: In view of ${\sf S}_k(n,\ell)={\sf G}_k(n,\ell,0)$ is

760: the constant term of $F_k(x)$, i.e.~$F_k(0)$, whence

761: \begin{equation}\label{E:y}

762: {\sf S}_k(n,\ell)=\sum_{b=0}^{(n-\ell)/2}\, (-1)^b\,

763:                   \binom{n-b}{b} f_k(n-2b,\ell) \ .

764: \end{equation}

765: It remains to prove eq~(\ref{E:sum}). Summing over all possible values

766: of isolated vertices, we get

767: \begin{align*}

768: S_{k}(n)&=\sum_{\ell=0}^{n}\sum_{b=0}^{(n-\ell)/2}(-1)^{b}{n-b

769: \choose b}f_{k}(n-2b,\ell) =

770: \sum_{b=0}^{\lfloor n/2\rfloor}(-1)^{b}{n-b \choose b}

771: \left\{\sum_{\ell=0}^{n-2b}f_{k}(n-2b,\ell)\right\}

772: \end{align*}

773: where $\sum_{\ell=0}^{n-2b}f_{k}(n-2b,\ell)$ is given by eq.~(\ref{E:ww2})

774: and the proof of the theorem is complete.

775: \end{proof}

776: \begin{tabular}{c|ccccccccccccccccccccc}

777: $n$ & \small{1} & \small{2} & \small{3} &\small{4} & \small{5} &

778: \small{6} & \small{7} & \small{8} & \small{9} & \small{10} &

779: \small{11} & \small{12} & \small{13} & \small{14} &

780: \small{15}\\

781: \hline $S_{3}(n)$ & \small{1} & \small{1} & \small{2} &\small{5} &

782: \small{13} &\small{36} & \small{105} & \small{321} & \small{1018} &

783: \small{3334} &\small{11216} &\small{38635} &\small{135835}

784: &\small{486337} &\small{1769500}\\

785: \end{tabular}

786: 

787: {\small {\bf Table 1.} The first $15$ numbers of $3$-noncrossing RNA

788: structures.}

789: 

790: A first implication of Theorem~\ref{T:cool1} is a new proof

791: for Waterman's formula \cite{Waterman:94a} for the number of

792: RNA secondary structures with exactly $k$ base pairs or

793: equivalently having $\ell=n-2k$ isolated vertices.

794: 

795: %%%

796: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

797: %%%

798: \begin{corollary}\label{C:k=2}

799: The number of RNA secondary structures having exactly $\ell$ isolated

800: vertices, ${\sf S}_2(n,\ell)$, is given by

801: \begin{equation}\label{E:Waterman-tree}

802: {\sf S}_2(n,\ell)  =

803:       \frac{2}{n-\ell}{\frac{n+\ell}{2} \choose \frac{n-\ell}{2} +1}

804:                     {\frac{n+\ell}{2}-1 \choose \frac{n-\ell}{2}-1} \  .

805: \end{equation}

806: Furthermore ${\sf S}_2(n,\ell)$ satisfies the recursion

807: \begin{equation}\label{E:rec2}

808: (n-\ell)(n-\ell+2)\cdot {\sf S}_{2}(n,\ell)\, -\, (n+\ell)(n+\ell-2)\cdot

809: {\sf S}_{2}(n-2,\ell)=0 \ .

810: \end{equation}

811: \end{corollary}

812: %%%

813: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

814: %%%

815: \begin{proof}

816: We actually give two independent proofs of eq~(\ref{E:Waterman-tree}):

817: the first being a direct computation based on eq.~(\ref{E:da}) and the

818: second using the recursion in eq.~(\ref{E:rec2}) derived by Zeilberger's

819: algorithm \cite{Zeilberger:96a}. Let $\frac{n-\ell}{2}=k$ we compute

820: \begin{align*}

821: {\sf S}_2(n,\ell)

822:        &=\sum_{b=0}^{(n-\ell)/2}(-1)^b\binom{n-b}{b}\binom{n-2b}{\ell}

823:                    \,  C_{\frac{n-\ell-2b}{2}}\\

824: &=\sum_{b=0}^{k}(-1)^{b}\frac{(n-b)!}{b!(n-2k)!}\cdot

825: \frac{1}{(k-b+1)!(k-b)!}\\

826: &=\frac{(n-k-1)!}{(n-2k)!\cdot

827: k!}\sum_{b=0}^{k}(-1)^{b}\frac{(n-b)!}{(k-b+1)!(n-k-1)!}

828: \frac{k!}{b!(k-b)!}\\

829: &=\frac{1}{n-k}{n-k \choose k}\sum_{b=0}^{k}(-1)^{b}{k \choose

830: b}{n-b \choose k-b+1}\\

831: &=(-1)^{k+1}\frac{1}{n-k}{n-k \choose k}\sum_{b=0}^{k}{k \choose

832: b}{k-n \choose k-b+1}\\

833: &=\frac{1}{n-k}{n-k \choose k+1}{n-k \choose k}\\

834: &=\frac{1}{k}{n-k \choose k+1}{n-k-1 \choose k-1}.

835: \end{align*}

836: As for the second proof we use ${\sf S}_2(n,\ell)=\sum_{b=0}^{(n-\ell)/2}

837: (-1)^b\binom{n-b}{b}f_2(n-2b,\ell)$ as

838: the input for Zeilberger's algorithm \cite{Zeilberger:96a} and obtain that

839: ${\sf S}_{2}(n,\ell)$ satisfies the recursion formula

840: \begin{equation}\label{E:rec21}

841: (n-\ell)(n-\ell+2)\cdot {\sf S}_{2}(n,\ell)-(n+\ell)(n+\ell-2)\cdot

842: {\sf S}_{2}(n-2,\ell)=0 \ .

843: \end{equation}

844: Using a bijection between RNA secondary structures and linear trees

845: Waterman computed in \cite{Waterman:94a} the number of RNA secondary

846: structures with exactly $h$ arcs, $s(n,h)$

847: \begin{equation}\label{E:water}

848: s(n,h)=\frac{1}{h}{n-h \choose h+1}{n-h-1 \choose h-1} \ .

849: \end{equation}

850: It follows by direct computation that ${\sf S}_2(n,n-2h)=s(n,h)=

851: \frac{1}{h}{n-h \choose h+1}{n-h-1 \choose h-1}$ satisfies the

852: recursion in eq.~(\ref{E:rec21}), from which we can conclude

853: ${\sf S}_2(n,\ell) =s(n,(n-\ell)/2)$.

854: \end{proof}

855: 

856: %%%

857: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

858: %%%

859: \begin{corollary}\label{C:k=3}

860: The number of $3$-noncrossing RNA structures having exactly $\ell$

861: isolated vertices, ${\sf S}_3(n,\ell)$, is given by

862: \begin{equation}\label{E:k=3ex}

863: {\sf S}_3(n,\ell)  =  \sum_{b=0}^{(n-\ell)/2}(-1)^b\binom{n-b}{b}

864:  \binom{n-2b}{\ell} \left[C_{\frac{n-\ell-2b}{2}}\,

865: C_{\frac{n-\ell-2b}{2}+2}-C_{\frac{n-\ell-2b}{2}+1}^2\right] \ .

866: \end{equation}

867: \end{corollary}

868: %%%

869: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

870: %%%

871: 

872: Using the expression of Corollary~\ref{C:k=3} for ${\sf S}_3(n,\ell)$ as

873: an input for Zeilberger's algorithm \cite{Zeilberger:96a} we derive

874: 

875: %%%

876: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

877: %%%

878: \begin{corollary}\label{C:recursion}

879:  The number of $3$-noncrossing RNA structures having

880: exactly $\ell$ isolated vertices, ${\sf S}_3(n,\ell)$, satisfies the

881: $4$-term recursion

882: \begin{align}

883: {\sf p}_1(n)\, {\sf S}_{3}(n-6,\ell)-{\sf p}_2(n)\, {\sf S}_{3}(n-4,\ell)-

884: {\sf p}_3(n){\sf S}_{3}(n-2,\ell)+{\sf p}_4(n)\,{\sf S}_{3}(n,\ell)=0 \ ,

885: \end{align}

886: where the coefficients ${\sf p}_1(n,\ell)$, ${\sf p}_2(n,\ell)$

887: ${\sf p}_3(n,\ell)$ and ${\sf p}_4(n,\ell)$ are given by

888: \begin{eqnarray*}

889: {\sf p}_1(n,\ell) & = &

890: \frac{1}{2}n(n-1)(n-10+\ell)(n-4+\ell)(n-8+\ell) \\

891: {\sf p}_2(n,\ell) & = & \frac{1}{2}n(n-3)(13n^{3}-126n^{2}+13n^{2}\ell-

892:               88n\ell+392n+3n\ell^{2}+216\ell-384-42\ell^{2}+3\ell^{3}) \\

893: {\sf p}_3(n,\ell) & = & (n-1)(\frac{1}{2}n-2)(13n^{3}-30n^{2}-13n^{2}\ell+

894:              8n+16n\ell+3n\ell^{2}+30\ell^{2}-72\ell-3\ell^{3}) \\

895: {\sf p}_4(n,\ell) & = &

896: (n-3)(\frac{1}{2}n-2)(n-\ell)(n-\ell+6)(n-\ell+4) \ .

897: \end{eqnarray*}

898: \end{corollary}

899: %%%

900: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

901: %%%

902: 

903: Theorem~\ref{T:cool1} immediately allows us to

904: derive the generating function for circular $k$-noncrossing RNA

905: structures.

906: Circular RNA structures are $k$-noncrossing digraphs without arcs of the

907: form $(1,n)$, representing molecular structures over circular sequences.

908: In circular sequences the arc $(n,1)$ is considered a bond of the primary

909: sequence and consequently does not occur as an arc in the corresponding

910: digraph representation. Suppose $k\ge 2$ and let $\mathcal{G}^{(c)}_{n,k}

911: (\ell,j)$ be the set of all $k$-noncrossing digraphs having exactly $\ell$

912: isolated points and exactly $j$ $1$-arcs, where a $1$-arc is an arc of the

913: form $(i,i+1)$, where $i$ is considered modulo $n$.

914: We set ${\sf G}^{(c)}_{k}(n,\ell,j)=\vert \mathcal{G}^{(c)}_{n,k}

915: (\ell,j)\vert$.

916: 

917: %%%

918: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

919: %%%

920: \begin{theorem}\label{T:circular}

921: Let $k\in\mathbb{N}$, $k\ge 2$, then the number of circular

922: $k$-noncrossing RNA structures, with exactly $\ell$ isolated vertices

923: ${\sf S}^{(c)}_k(n,\ell)$, is given by

924: \begin{equation}\label{E:dacg}

925: {\sf S}^{(c)}_k(n,\ell) = \sum_{b=0}^{(n-\ell)/2}

926: (-1)^b \left[\binom{(n-2)-(b-1)}{b-1}+\binom{n-b}{b}\right]

927: f_k(n-2b,\ell)  \  .

928: \end{equation}

929: where $\lambda^{(c)}(n,b)(1,0)=0$, $\lambda^{(c)}(n,b)(1,1)=1$,

930: $\lambda^{(c)}(2,0)=0$ and  $\lambda^{(c)}(2,2)=1$.

931: \begin{equation}\label{E:sumcirc}

932: {\sf S}^{(c)}_k(n)

933: =\sum_{b=0}^{\lfloor n/2\rfloor}(-1)^{b}\left[\binom{(n-2)-(b-1)}{b-1}+

934: \binom{n-b}{b}\right]\left\{\sum_{\ell=0}^{n-2b}f_{k}(n-2b,\ell)\right\}

935: \end{equation}

936: where $\sum_{\ell=0}^{n-2b}f_{k}(n-2b,\ell)$ is given by eq.~{\rm

937: (\ref{E:ww2})}.

938: \end{theorem}

939: %%%

940: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

941: %%%

942: \begin{proof}

943: For circular RNA structures the $1$-arcs are considered modulo $n$.

944: Again we derive a family $\mathcal{F}$ of $\mathcal{G}_{n,k}$-digraphs,

945: having exactly $\ell$ isolated points and at least $b$ $1$-arcs.

946: We select {\sf (a)} $b$ $1$-arcs, and {\sf (b)} an arbitrary

947: $k$-noncrossing digraph with exactly $\ell$ isolated points over the

948: remaining $n-2b$ vertices. In complete analogy we derive that each element

949: $\theta\in \mathcal{F}$ is contained in $\mathcal{G}^{(c)}_{n,k}(\ell,j)$

950: for some $j\ge b$.

951: Let $\lambda^{(c)}(n,b)$, denote the number of ways to select $b$

952: $1$-arcs over $\{1,\dots,n\}$ including the arc $(n,1)$. Then

953: $\lambda^{(c)}(n,b)$ is given by

954: \begin{equation}\label{E:circular}

955: \lambda^{(c)}(n,b)=

956: \binom{(n-2)-(b-1)}{b-1}+\binom{n-b}{b} \ ,

957: \end{equation}

958: where $\lambda^{(c)}(n,b)(1,0)=0$, $\lambda^{(c)}(n,b)(1,1)=1$,

959: $\lambda^{(c)}(2,0)=0$ and  $\lambda^{(c)}(2,2)=1$.

960: Indeed, either the arc $(n,1)$ is selected in which case we are left with

961: exactly $\binom{(n-2)-(b-1)}{b-1}$ ways to select the remaining $1$-arcs

962: or $(n,1)$ is not selected, in which case according to

963: Theorem~\ref{T:cool1}

964: there are exactly $\binom{n-b}{b}$ ways to select the $1$-arcs. Therefore

965: we obtain

966: \begin{equation}\label{E:for}

967: \sum_{j\ge b}\binom{j}{b}\, {\sf G}^{(c)}_k(n,\ell,j)=

968: \left[\binom{(n-2)-(b-1)}{b-1}+\binom{n-b}{b}\right]

969: \, f_k(n-2b,\ell) \ .

970: \end{equation}

971: In complete analogy to the argument in Theorem~\ref{T:cool1} we can

972: conclude

973: \begin{equation}

974: {\sf S}^{(c)}_k(n,\ell)=\sum_{b=0}^{(n-\ell)/2}\, (-1)^b\,

975:    \left[\binom{(n-2)-(b-1)}{b-1}+\binom{n-b}{b}\right]f_k(n-2b,\ell) \ .

976: \end{equation}

977: Eq.~(\ref{E:sumcirc}) follows analogously and the proof of the theorem is

978: complete.

979: \end{proof}

980: 

981: %%%

982: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

983: %%%

984: \section{Restricted RNA structures}\label{S:2-arcs}

985: %%%

986: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

987: %%%

988: 

989: We now generalize the ideas in Section~\ref{S:RNA} for the enumeration

990: of restricted RNA structures.

991: A restricted RNA structure is an RNA structure without any $2$-arcs,

992: i.e.~arcs of the form $(i,i+2)$.

993: In this case we need the condition $k>2$

994: instead of $k\ge 2$, since our construction can produce $2$-sets of

995: mutually crossing arcs.

996: Let $\mathcal{G}_{n,k}(\ell,j_1,j_2)$ be the set of all $k$-noncrossing

997: digraphs having exactly $\ell$ isolated points and exactly $j_1$ and

998: $j_2$ $1$-and $2$-arcs. We set

999: $

1000: {\sf G}_{k}(n,\ell,j_1,j_2)=\vert \mathcal{G}_{n,k}(\ell,j_1,j_2)\vert

1001: $.

1002: In particular we have ${\sf G}_k(n,\ell,0,0)={\sf S}_k^{(r)}(n,\ell)$.

1003: 

1004: %%%

1005: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

1006: %%%

1007: \begin{theorem}\label{T:cool2}

1008: Let $k\in\mathbb{N}$, $k>2$. Then the numbers of restricted RNA structures

1009: ${\sf S}_k^{(r)}(n,\ell)$ and ${\sf S}^{(r)}_k(n)$ are given by

1010: \begin{eqnarray}\label{E:da2}

1011: {\sf S}_k^{(r)}(n,\ell) & = &

1012: \sum_{b_{1}\ge 0,b_{2}\ge 0}(-1)^{b_{1}+b_{2}}

1013: \lambda(n,b_{1},b_{2})f_{k}(n-2(b_{1}+b_{2}),\ell) \\

1014: \label{E:da3}

1015: {\sf S}_k^{(r)}(n) & = & \sum_{b_{1}\ge 0,b_{2}\ge 0}^{\lfloor n/2\rfloor}

1016: (-1)^{b_{1}+b_{2}} \lambda(n,b_1,b_2)

1017: \left\{\sum_{\ell=0}^{n-2(b_1+b_2)}f_{k}(n-2(b_{1}+b_{2}),\ell)\right\}\ .

1018: \end{eqnarray}

1019: Here $\lambda(n,b_1,b_2)$ satisfies the recursion

1020: \begin{equation}\label{E:hh}

1021: \lambda(n,b_1,b_2)  =  \lambda(n-2,b_1-1,b_2)

1022: +\lambda(n-1,b_1,b_2) +\lambda(n-4,b_1,b_2-2) +

1023: \lambda(n-3,b_1,b_2-1)

1024: \end{equation}

1025: and the initial conditions for eq.~{\rm (\ref{E:hh})} are

1026: $\lambda(n,0,0)=1$,

1027: $\lambda(n,b_{1},0)={n-b_{1} \choose b_{1}}$, $\lambda(n,0,b_{2})=

1028: \gamma(n,b_{2})$ and $\gamma(n,1)=0$ for $n=1$, $\gamma(n,1)=n-2$ for

1029: $n\ge 2$ and $\gamma(n,2)=0$ for $n=2,3$.

1030: \end{theorem}

1031: %%%

1032: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

1033: %%%

1034: \begin{proof}

1035: Suppose $\lambda(n,b_1,b_2)$ is the number of ways to

1036: select exactly $b_1$ $1$-arcs and $b_2$ $2$-arcs over $\{1,\dots,n\}$

1037: vertices.\\

1038: {\it Claim.} $\lambda(n,b_1,b_2)$ satisfies the recursion of

1039: eq.~(\ref{E:hh}) with the respective initial conditions, and we have

1040: \begin{equation}\label{E:form2}

1041: \sum_{j_1\ge b_1,j_2\ge b_2}\binom{j_1}{b_1}\,\binom{j_2}{b_2}\,

1042: {\sf G}_k(n,\ell,j_1,j_2)=

1043: \lambda(n,b_1,b_2)\, f_k(n-2(b_1+b_2),\ell) \ .

1044: \end{equation}

1045: In analogy to the proof of Theorem~\ref{T:cool1} we derive a family

1046: $\mathcal{F}$ of $\mathcal{G}_{n,k}$-digraphs, having exactly $\ell$

1047: isolated points and at least $b_1$ and $b_2$ $1$-arcs and $2$-arcs,

1048: respectively.

1049: We first prove that this construction generates elements of

1050: $\mathcal{G}_{n,k}(\ell,j_1,j_2)$ and then express

1051: $\vert\mathcal{F}\vert$ via the numbers ${\sf G}_k(n,\ell,j_1,j_2)$.

1052: We select {\sf (a)} $b_1$ $1$-arcs

1053: and $b_2$ $2$-arcs and {\sf (b)} an arbitrary $k$-noncrossing digraph

1054: over the remaining $n-2(b_1+b_2)$ vertices with exactly $\ell$ isolated

1055: points. Let $\mathcal{F}$ be the family of digraphs obtained this way. \\

1056: {\it Claim $1$.} Each element $\theta\in \mathcal{F}$ is contained in

1057: $\mathcal{G}_{n,k}(\ell,j_1,j_2)$ for some $j_1\ge b_1$ and $j_2\ge b_2$.\\

1058: To prove this we observe that any $1$-arc or $2$-arc can only cross at

1059: most one other arc. Therefore $1$-arcs and $2$-arcs cannot be contained

1060: in a set of more than $2$-mutually crossing arcs. As a result, for $k>2$

1061: we generate digraphs that are $k$-noncrossing. Clearly $\theta$ has

1062: exactly

1063: $\ell$ isolated vertices and in step {\sf (b)} we potentially derive

1064: additional $1$-arcs and $2$-arcs, whence $j_1\ge b_1$ and $j_2\ge b_2$,

1065: respectively.\\

1066: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

1067: {\it Claim $2$.}

1068: \begin{equation}\label{E:fam2}

1069: \vert \mathcal{F}\vert =  \lambda(n,b_1,b_2)\, f_k(n-2(b_1+b_2),\ell)\ .

1070: \end{equation}

1071: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

1072: We prove that the number of ways to select $1$ and $2$-arcs

1073: satisfies the recursion in eq~(\ref{E:hh}) by induction on $n$.

1074: For the induction step we distinguish the following cases:\\

1075: {\sf Case 1.} The arc $(1,2)$ is selected. Then we have

1076: $\lambda(n-2,b_1-1,b_2)$ ways to select $(b_1-1)$ $1$-arcs and $b_2$

1077: $2$-arcs over the vertices $\{3,\dots,n\}$.\\

1078: {\sf Case 2.} The arc $(1,2)$ not selected. Then we distinguish the

1079: scenarios: $(1,3)$ is selected and $(1,3)$ is not selected. In the latter

1080: case we have $\lambda(n-1,b_1,b_2)$ ways to choose $b_1$ $1$-arcs and

1081: $b_2$ $2$-arcs over the vertices $\{2,\dots,n\}$.

1082: Suppose $(1,3)$ is selected. Then we have either that $(2,4)$ is selected,

1083: in which case we can select the remaining $b_1$ $1$-arcs and $b_2$

1084: $2$-arcs over $\{5,\dots,n\}$ in exactly $\lambda(n-4,b_1,b_2-2)$

1085: different ways.

1086: In case $(2,4)$ is not selected we can freely choose $b_1$ $1$-arcs and

1087: $(b_2-1)$ $2$-arcs over $\{4,\dots,n\}$ i.e.~there are

1088: $\lambda(n-3,b_1,b_2-1)$ ways. Therefore we derive the recursion

1089: $$

1090: \lambda(n,b_1,b_2) =  \lambda(n-2,b_1-1,b_2)+

1091: \lambda(n-1,b_1,b_2) +\lambda(n-4,b_1,b_2-2)+\lambda(n-3,b_1,b_2-1)  \,  .

1092: $$

1093: As for the intial conditions, we have are $\lambda(n,0,0)=1$,

1094: $\lambda(n,b_{1},0)={n-b_{1} \choose b_{1}}$, $\lambda(n,0,b_{2})=

1095: \gamma(n,b_{2})$ and $\gamma(n,1)=0$ for $n=1$, $\gamma(n,1)=n-2$ for

1096: $n\ge 2$ and $\gamma(n,2)=0$ for $n=2,3$.

1097: It remains to select an arbitrary $k$-noncrossing digraph with $\ell$

1098: isolated vertices over $n-2(b_1+b_2)$ vertices. According to

1099: Lemma~\ref{L:walks} the

1100: latter number is given by $f_k(n-2(b_1+b_2),\ell)$,

1101: whence eq.~(\ref{E:fam2}) and Claim $2$ is proved.

1102: In view of the fact that any of the $k$-noncrossing digraphs over

1103: $n-2(b_1+b_2)$ vertices can introduce additional $1$-arcs or

1104: $2$-arcs, we set

1105: $$

1106: \mathcal{F}(j_1,j_2)=\{\theta\in\mathcal{F} \mid \theta\ \text{\rm has

1107:                      exactly $j_1$ $1$-arcs and $j_2$ $2$-arcs} \} \ .

1108: $$

1109: Obviously, we have the partition $\mathcal{F}=\dot\bigcup_{j_1 \ge b_1,

1110: \,j_2\ge b_2}\mathcal{F}(j_1,j_2)$. Suppose $\theta \in

1111: \mathcal{F}(j_1,j_2)$.

1112: According to Claim $1$, $\theta\in \mathcal{G}_{n,k}(\ell,j_1,j_2)$

1113: and furthermore $\theta$ occurs with multiplicity $\binom{j_1}{b_1}$

1114: $\binom{j_2}{b_2}$ in $\mathcal{F}$ since by construction any

1115: $b_1$-element subset of the $j_1$ $1$-arcs and

1116: $b_2$-element subset of the $j_2$ $2$-arcs is counted respectively in

1117: $\mathcal{F}$. Therefore we have

1118: \begin{equation} \label{E:well2}

1119: \vert \mathcal{F}(j_1,j_2)\vert = \binom{j_1}{b_1}\binom{j_2}{b_2}

1120: {\sf G}_k(n,\ell,j_1,j_2)

1121: \end{equation}

1122: and

1123: \begin{eqnarray*}

1124: \sum_{j_1 \ge b_1,\, j_2\ge b_2}

1125: \binom{j_1}{b_1}\binom{j_2}{b_2}{\sf G}_k(n,\ell,j_1,j_2) & = &

1126: \sum_{j_1 \ge b_1,\, j_2\ge b_2} \vert \mathcal{F}(j_1,j_2)\vert \\

1127: & = &  \lambda(n,b_1,b_2)f_k(n-2(b_1+b_2),\ell) \ .

1128: \end{eqnarray*}

1129: We next set $F_{k}(x,y)=\sum_{j_{1}\ge 0}\sum_{j_{2} \ge 0}{\sf

1130: G}_{k}(n,\ell,j_{1},j_{2})x^{j_{1}}y^{j_{2}}$.

1131: Taking the $b_1$-th and $b_2$-th derivatives w.r.t.~$x$ and $y$ we obtain

1132: \begin{eqnarray}\label{E:u2}

1133: \frac{1}{b_1!}\frac{1}{b_2!} F_k^{(b_1,b_2)}(1) =

1134: \sum_{j_1\ge b_1,\,j_2\ge b_2}

1135: \binom{j_1}{b_1}\binom{j_2}{b_2}{\sf G}_k(n,\ell,j_1,j_2)\, 1^{j_1-b_1}

1136: 1^{j_2-b_2}

1137:  \ .

1138: \end{eqnarray}

1139: Then we have

1140: \begin{align*}

1141: \sum_{j_{1},j_{2} \ge 0}

1142: {\sf G}_{k}(n,\ell,j_{1},j_{2})x^{j_{1}}y^{j_{2}}

1143: &=\sum_{b_{1}\ge 0,b_{2}\ge 0}\left[\sum_{j_{1}\ge b_{1},j_{2}\ge

1144: b_{2}}{j_{1} \choose b_{1}}{j_{2} \choose

1145: b_{2}}{\sf G}_{k}(n,\ell,j_{1},j_{2})\right](x-1)^{b_{1}}(y-1)^{b_{2}}\\

1146: &=\sum_{b_{1}\ge 0,b_{2}\ge 0}

1147: \lambda(n,b_{1},b_{2})\, f_k(n-2(b_1+b_2),\ell)\,(x-1)^{b_{1}}\,

1148: (y-1)^{b_{2}} \ .\\

1149: \end{align*}

1150: By construction ${\sf G}(n,\ell,0,0)$ is the constant term of the

1151: $F_{k}(x,y)$. That is, the number of k-noncrossing RNA structures

1152: with $\ell$ isolated vertices and no 2-arcs is given by

1153: \begin{equation}

1154: {\sf G}(n,\ell,0,0)=

1155: \sum_{b_{1}\ge 0,b_{2}\ge 0}(-1)^{b_{1}+b_{2}}

1156: \lambda(n,b_{1},b_{2})f_{k}(n-2(b_{1}+b_{2}),\ell)

1157: \end{equation}

1158: and taking the sum over all $\ell$ eq.~(\ref{E:da3}) follows

1159: $$

1160: {\sf S}_k^{(r)}(n)  =  \sum_{b_{1}\ge 0,b_{2}\ge 0}^{\lfloor n/2\rfloor}

1161: (-1)^{b_{1}+b_{2}} \lambda(n,b_1,b_2)

1162: \left\{\sum_{\ell=0}^{n-2(b_1+b_2)}f_{k}(n-2(b_{1}+b_{2}),\ell)\right\}

1163: \ ,

1164: $$

1165: where

1166: $\left\{\sum_{\ell=0}^{n-2(b_1+b_2)}f_{k}(n-2(b_{1}+b_{2}),\ell)\right\}$

1167: is given by eq.~(\ref{E:ww2}) and the proof of the theorem is complete.

1168: \end{proof}

1169: 

1170: \begin{tabular}{c|ccccccccccccccccccccc}

1171: $n$ & \small{1} & \small{2} & \small{3} &\small{4} & \small{5} &

1172: \small{6} & \small{7} & \small{8} & \small{9} & \small{10} &

1173: \small{11} & \small{12} & \small{13} & \small{14} &

1174: \small{15}\\

1175: \hline $S_{3}^{r}(n)$ & \small{1} & \small{1} & \small{1} &\small{2}

1176: & \small{5} &\small{14} & \small{40} & \small{119} & \small{364} &

1177: \small{1145} &\small{3688} &\small{12139} &\small{40734}

1178: &\small{139071} &\small{482214}\\

1179: \end{tabular}

1180: 

1181: {\small {\bf Table 3.} The first $15$ numbers of $3$-noncrossing restricted

1182:         RNA structures.}

1183: 

1184: 

1185: %%%

1186: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

1187: %%%

1188: \section{Appendix}\label{S:appendix}

1189: %%%

1190: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

1191: %%%

1192: {\bf Proof of Theorem~\ref{T:tableaux}.}

1193: Suppose we have two shapes $\mu^{i}\subsetneq \mu^{i-1}$ and $T_{i-1}$

1194: is a standard Young tableau of shape $\mu^{i-1}$. We first observe that

1195: there exists a unique $j$ and a unique $T_i$ such that $T_{i-1}$ is

1196: obtained from $T_{i}$ by row-inserting $j$ with the RSK algorithm. \\

1197: Suppose $\mu^{i-1}$ differs from $\mu^{i}$ in the first row. Then

1198: $j$ is the element at the end of the first row in $T_{i-1}$.

1199: Otherwise suppose $\ell$ is the row of the square being removed from

1200: $T_{i-1}$. Remove the square and insert its element $x$ into the

1201: $(\ell-1)$-th row at precisely the position, where the removed element

1202: $y$ would push it down via the RSK-algorithm. That is $y$ is maximal

1203: subject to $y<x$. Since each column is strictly increasing $y$

1204: always exists. Iterating this process results in exactly one element

1205: $j$ being removed from $T_i$ and a new filling of $\mu_{i-1}$,

1206: i.e.~a unique tableau $T_{i-1}$. By construction, inserting

1207: $j$ with the RSK algorithm produces $T_{i-1}$.\\

1208: {\it Claim $1$.} There exists a bijection between the set of oscillating

1209: tableaux of length $n$ and digraphs with vertices of degree $\le 1$.\\

1210: Given an oscillating tableau

1211: $(\mu^i)_{i=0}^n$ ($\mu^i$ differs from $\mu^{i-1}$ by at most one square),

1212: we recursively define a sequence

1213: $(G_{0},T_{0}),(G_{1},T_{1}),\ldots,(G_{n},T_{n})$,

1214: where $G_i$ is a digraph and $T_{i}$

1215: is a standard Young tableau. We define $G_0$ to be the digraph with

1216: empty edge-set and $T_{0}$ to be the empty standard

1217: Young tableau.

1218: The tableau $T_{i}$ is obtained from $T_{i-1}$ and

1219: the digraph $G_{i}$ is obtained from $G_{i-1}$ by the following

1220: procedure: \\

1221: {\sf 1.} {\sf (Insert origins)}

1222:    For $\mu^{i}\supsetneq\mu^{i-1}$, then

1223:    $T_{i}$ is obtained from $T_{i-1}$ by adding the entry $i$ in the

1224:    square $\mu^{i}\backslash \mu^{i-1}$.\\

1225: {\sf 2.} {\sf (Isolated vertices)}

1226: For $\mu^{i} = \mu^{i-1}$ then set $T_i=T_{i-1}$\\

1227: {\sf 3.} {\sf (Remove origins)} For $\mu^{i}\subsetneq \mu^{i-1}$, then

1228: let $T_{i}$ be the unique standard Young tableau of shape $\mu^{i}$

1229: and $j$ be the unique number such that $T_{i-1}$ is obtained from $T_{i}$

1230: by row-inserting $j$ with the RSK algorithm. Then set

1231: $E_{G_{i}}=E_{G_{i-1}}\cup \{(j,i)\}$.\\

1232: Obviously, $G_n$ is a digraph, and the set of $i$ where $\mu^{i} =

1233: \mu^{i-1}$ equals the set of isolated vertices of $G_n$.

1234: By construction each entry $j$ is removed exactly once whence no edges

1235: of the form $(j,i)$ and $(j,i')$ can be obtained. Therefore $G_n$ has

1236: degree $\le 1$ and we have a well defined mapping

1237: \begin{eqnarray*}

1238: \beta\colon

1239: \{(\mu_i)_{i=0}^n\mid (\mu_i)_{i=0}^n\,\text{\rm is an oscillating

1240: tableau} \} \longrightarrow \{G_n\mid G_n \, \text{\rm is a digraph with

1241: degree $\le 1$}\} \ .

1242: \end{eqnarray*}

1243: It is clear from the procedure that $G_n$ is a labeled graph and

1244: $\beta$ is injective. To prove surjectivity we observe that each

1245: digraph $G_n$ induces an oscillating tableau as follows. We set

1246: $\mu_{G_n}^n=\varnothing$ and $T_n=\varnothing$. Starting from

1247: vertex $i=n,n-1,\dots, 1,0$ we derive a sequence of Young tableaux

1248: $(T_n,T_{n-1},\dots,T_0)$ as follows:\\

1249: {\sf I.} If $i$ is an terminus of an $G_n$-arc $(j,i)$ add $j$ via the

1250:          RSK-algorithm to $T_{i}$ set

1251:          $\mu_{G_n}^{i-1}\supsetneq \mu_{G_n}^{i}$ to be the

1252:          shape of $T_{i-1}$ (corresponds to {\sf (3)})\\

1253: {\sf II.} If $i$ is an isolated $G_n$-vertex set $\mu_{G_n}^{i-1}=

1254:          \mu_{G_n}^{i}$ (corresponds to {\sf (2)})\\

1255: {\sf III.} If $i$ is the origin of an $G_n$-arc $(i,k)$ let

1256:          $\mu_{G_n}^{i-1}\subsetneq \mu_{G_n}^{i}$ be the shape of

1257:          $T_{i-1}$,

1258:          the standard Young tableau obtained by removing the square

1259:          containing $i$ (corresponds to {\sf (1)}).   \\

1260: Then we have $\beta((\mu_{G_n})_0^n)=G_n$, whence $\beta$ is surjective.\\

1261: {\it Claim $2$.} $G_n$ is $k$-noncrossing if and only if all shapes

1262: $\mu^i$ in the oscillating tableau have less than $k$ rows.\\

1263: From Claim 1 we know $\beta^{-1}(G_{n})=(\varnothing=\mu^{0},

1264: \mu^{1},\ldots \mu^{n}=\varnothing)$, so it suffices to prove that

1265: the maximal number of rows in the shape set $\beta^{-1}(G_{n})$ is

1266: less than $k$. First we observe that the arcs $(i_{1},j_{1}),\ldots

1267: (i_{\ell},j_{\ell})$ form a $\ell$-crossing of $G_{n}$ if and only

1268: if there exists a tableau $T_{i}$ such that elements

1269: $i_{1},i_{2},\ldots i_{\ell}$ are in the $\ell$ squares of $T_{i}$

1270: and being deleted in increasing order $i_{1}<i_{2}<\ldots i_{\ell}$

1271: afterwards. Next, we will obtain a permutation $\pi_{i}$ from the

1272: entries in each tableau $T_{i}$ recursively as follows:\\

1273:  {\sf 1.} If $T_{i-1}$ is obtained from $T_{i}$ by row-inserting $j$

1274:           with the RSK algorithm, then $\pi_{i-1}=\pi_{i}j$. \\

1275: {\sf 2.} If $T_i=T_{i-1}$, then $\pi_{i}=\pi_{i-1}$.\\

1276: {\sf 3.} If $T_{i-1}$ is obtained from $T_{i}$ by deleting the entry

1277:          $i$, then $\pi_{i-1}$ is obtained from $\pi_{i}$ by deleting

1278:          $i$.\\

1279: If $\pi=r_{1}r_{2}\ldots r_{t}$, then the entries being deleted

1280: afterwards are in the order $r_{t},\ldots r_{2},r_{1}$.\\

1281: Using the RSK algorithm w.r.t.~the permutation $\pi_{i}$, the resulting

1282: row-inserting Young tableau is exactly $T_{i}$. We prove this by

1283: induction in reverse order of the oscillating tableau. It is trivial

1284: for the case $i=n$. Suppose it holds for $j$, $1\le j\le n$. Consider

1285: the above three cases: inserting an element, doing nothing and deleting

1286: an element. In the first case, the assertion is implied the RSK

1287: algorithm in the construction of the oscillating tableau. In the second

1288: case, it holds by the induction hypothesis on step $j$.\\

1289: Now it remains to consider the third case, that is, removing the

1290: entry from $T_{j}$ to get $T_{j-1}$. Write $\pi_{j}=x_{1}x_{2}\ldots

1291: x_{p}jy_{1}y_{2}\ldots y_{q}$ and $\pi_{j-1}=x_{1}x_{2}\ldots

1292: x_{p}y_{1}y_{2}\ldots y_{q}$. In view of step {\sf 3} $j$ is larger than

1293: elements $x_{1},x_{2},\ldots,x_{p},y_{1},\ldots y_{q}$. We need to

1294: prove that the insertion tableau $S_{j-1}$ of $\pi_{j-1}$ by the RSK

1295: algorithm is exactly the same as deleting the entry $j$ in $T_{j}$.

1296: We proceed by induction on $q$. In the case $q=0$, $T_{j}$ is

1297: obtained from $T_{j-1}$ by adding $j$ at the end of the first row.

1298: Suppose the assertion holds for $q-1$, that is

1299: $S_{j-1}(x_{1}x_{2}\ldots x_{p}y_{1}y_{2}\ldots

1300: y_{q-1})=S_{j}(x_{1}x_{2}\ldots x_{p}jy_{1}y_{2}\ldots

1301: y_{q-1})\setminus\text{\rm \fbox{$\,j\,$}}$.

1302: Consider inserting $y_{q}$ into $S_{j-1}$,

1303: via the RSK algorithm.

1304: If the insertion track path never touches the position of $j$, then

1305: $S_{j-1}(x_{1}x_{2}\ldots x_{p}y_{1}y_{2}\ldots

1306: y_{q-1}y_{q})=S_{j}(x_{1}x_{2}\ldots x_{p}jy_{1}y_{2}\ldots

1307: y_{q-1}y_{q})\setminus\text{\rm \fbox{$\,j\,$}}$. Otherwise, if

1308: the insertion path

1309: touched $j$ and pushed $j$ into the next row, then since $j$ is greater

1310: than any other entry, $j$ must be moved to the end of next row and the

1311: push process stops. Accordingly, the

1312: insertion path in $S_{j-1}(x_{1}x_{2}\ldots x_{p}y_{1}y_{2}\ldots

1313: y_{q-1})$ is the

1314: same path as in $S_{j}(x_{1}x_{2}\ldots x_{p}jy_{1}y_{2}\ldots

1315: y_{q-1})$ except the last step moving $j$ to a new position $j$, so

1316: deleting $j$ will get $S_{j-1}(x_{1}x_{2}\ldots

1317: x_{p}y_{1}y_{2}\ldots y_{q-1}y_{q})=S_{j}(x_{1}x_{2}\ldots

1318: x_{p}jy_{1}y_{2}\ldots y_{q-1}y_{q})\setminus\text{\rm \fbox{$\,j\,$}}$.

1319: According to

1320: Schensted's Theorem, for any permutation $\pi$, assume $A$ is the

1321: corresponding insertion Young tableau by using the RSK algorithm on

1322: $\pi$. Then the length of the longest decreasing subsequences of

1323: $\pi$ is the number of rows in $A$, whence the assertion. \\

1324: Now we can prove Claim $2$. A diagraph is a $\ell$-crossing

1325: if and only if there exists a $\pi_{i}$ which has decreasing

1326: subsequence of length $\ell$. And the insertion Young tableau of

1327: $\pi_{i}$ is exactly the same with the labeled oscillating tableau

1328: $T_{i}$. According to Schensted's theorem, $\pi$ has a decreasing

1329: sequence of length $\ell$ if and only if rows of $T_{i}$ is $\ell$.\\

1330:  {\it Claim $3$.} There is a bijection

1331: between oscillating tableaux with at most $k-1$ rows of length $n$

1332: and walks with steps $\pm e_i,0$ which stay in the interior of

1333: $C_{0}$ starting and ending at

1334: $(k-1,k-2,\ldots,1)$.\\

1335: This bijection is obtained by setting for $1\le \ell\le k-1$, $x_\ell$

1336: to be the length of the $\ell$-th row. By definition of standard Young

1337: tableaux, we have $\lambda_{1}\ge \lambda_{2}\ge \ldots \lambda_{n}$

1338: i.e.~the length of each row is weakly decreasing. This property also

1339: characterizes walks that stay within the Weyl-chamber $C_0$, i.e.~where

1340: we have $x_{1}> x_{2}\ldots > x_{k-1}>0$ since a walk from

1341: $(k-1,\ldots 2,1)$ to itself in the interior of $C_{0}$ is a translation

1342: of a walk from the origin to itself in the region $x_{1}\ge x_{2}\ldots

1343: \ge x_{k-1}\ge 0$.

1344: In an oscillating tableau $\mu^i$ differs from $\mu^{i-1}$ by at most one

1345: square and adding or deleting a square in the $\ell$-th row or doing

1346: nothing corresponds to steps $\pm e_{\ell}$ and $0$, respectively.

1347: Since the oscillating tableau is of empty shape, we have

1348: walks from the origin to itself, whence Claim $3$ follows and the proof of

1349: the Theorem is complete. $\square$\\

1350: 

1351: %%%

1352: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

1353: %%%

1354: 

1355: {\bf Proof of Lemma~\ref{L:touch}.}

1356: To prove the lemma we can w.l.o.g.~assume $C=C_0=\{(x_1,\dots,x_{k-1})\mid

1357: x_1> x_2 >\dots> x_{k-1}>0 \}$. Then the assertion is that every walk

1358: having

1359: steps $\pm e_i,0$ starting at $a=(k-1,k-2,\dots,1)$ that crosses from

1360: inside ${C_0}$ into outside ${C_0}$ intersects one of the sub-spaces

1361: $\langle e_1\rangle$ or $\langle e_j-e_{j-1}\rangle$ for $2\le j\le k-1$.

1362: This is correct since to leave $C$ implies that there exists some $i$

1363: such that $x_i\le x_{i+1}$. Let $s_j$ be minimal

1364: w.r.t.~$a+\sum_h^{j+1}s_{h}\not\in C_0$. Since we have steps $\pm e_i,0$

1365: we conclude $x_{k-1}=0$ or $x_j=x_{j-1}$ for some $2\le j\le k-1$, whence

1366: the lemma. $\square$

1367: %%%

1368: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

1369: %%%

1370: 

1371: {\bf Proof of Theorem~\ref{T:reflect}.}

1372: Totally order the roots of $\Delta$. Let $\Gamma_n^-(a,b)$ be the

1373: number of walks $\gamma$ from $a$ to $b$, $a,b\in\mathbb{Z}^{k-1}$ of

1374: length $n$ using the steps $s$, $s\in \{\pm e_i,0\}$ such that

1375: $\langle \gamma(s_r),\alpha\rangle=0$ for some $\alpha\in \Delta$

1376: (i.e.~the walk intersects with the subspace $\langle

1377: \alpha\rangle$).

1378: According to Lemma~\ref{L:touch} every walk that crosses

1379: from inside ${C}$ into outside ${C}$ touches a wall from which we

1380: can draw two conclusions:

1381: \begin{eqnarray}\label{E:split}

1382: \Gamma_n(a,b) & = &\Gamma_n^{+}(a,b)+\Gamma_n^-(a,b) \\

1383: \text{\rm $\beta\neq {\sf id}$}\quad \Longrightarrow \quad

1384: \Gamma_n(\beta(a),b) & = &\Gamma_n^-(\beta(a),b) \ .

1385: \end{eqnarray}

1386: {\it Claim.} $\sum_{\beta\in {\sf B}_{k-1}}(-1)^{\ell(\beta)} \,

1387: \Gamma_n^-(\beta(a),b)=0$.\\

1388: Let $(s_1,\dots,s_n)$ be a walk from $\beta(a)$ to $b$.

1389: By assumption there

1390: exists some step $s_r$ at which we have $(\gamma_{\beta(a),b}(s_r),\alpha)

1391: =0$, for $\alpha\in \Delta$. Let $\alpha^*$ be the largest root for which

1392: we have $(\gamma_{\beta(a),b}(s_r),\alpha^*)=0$ and

1393: $\beta_{\alpha^*}(x)=x-\frac{2\langle \alpha^*,

1394: x\rangle }{\langle \alpha^*,\alpha^*\rangle}\alpha^*$

1395: its associated reflection (eq.~(\ref{E:Co})).

1396: We consider the walk

1397: \begin{equation}

1398: (\beta_{\alpha^*}(s_1),\dots,\beta_{\alpha^*}(s_r),s_{r+1},\dots,s_n)

1399: \end{equation}

1400: Now by definition $(\beta_{\alpha^*}(s_1),\dots,\beta_{\alpha^*}(s_r),

1401: s_{r+1},\dots,s_n)$ starts at $(\beta_{\alpha^*}\circ \beta)(a)$ and has

1402: sign $(-1)^{\ell(\beta)+1}$ since $\ell(\beta)+1=

1403: \ell(\beta_{\alpha^*}\circ \beta)$.

1404: Therefore to each

1405: element $\gamma_{\beta(a),b}$ of $\Gamma_n^-(\beta(a),b)$ having sign

1406: $(-1)^{\ell(\beta)}$ there exits a

1407: $\gamma_{\beta_{\alpha^*}\beta(a),b}\in\Gamma_n^-(\beta_{\alpha^*}

1408: \beta(a),b)$ with sign $(-1)^{\ell(\beta)+1}$ and the claim follows.

1409: We immediately derive

1410: \begin{eqnarray*}

1411: \sum_{\beta\in {\sf B}_{k-1}}(-1)^{\ell(\beta)}\,

1412: \Gamma_n(\beta(a),b) & = &

1413: \Gamma_n(a,b) +

1414: \sum_{\beta\in {\sf B}_{k-1},\beta\neq {\sf id}} (-1)^{\ell(\beta)} \,

1415: \underbrace{\Gamma_n(\beta(a),b)}_{=\Gamma_n^-(\beta(a),b)}\\

1416: & = & \Gamma_n^+(a,b) + \underbrace{\Gamma_n^{-}(a,b)+\sum_{\beta\in

1417: {\sf B}_{k-1},\beta\neq {\sf id}}

1418:                            (-1)^{\ell(\beta)} \, \Gamma_n^-(\beta(a),b)}_{

1419:  \sum_{\beta\in {\sf B}_{k-1}}(-1)^{\ell(\beta)} \,

1420: \Gamma_n^-(\beta(a),b)=0} \ ,

1421: \end{eqnarray*}

1422: whence the theorem. $\square$

1423: %%%

1424: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

1425: %%%

1426: 

1427: {\bf Proof of Lemma~\ref{L:walks}.}

1428: Let $u_i$, $1\le i\le k-1$ be transcendent variables and

1429: $u=(u_i)_1^{k-1}$.

1430: We define

1431: $

1432: {u}^{b-a}=\prod_{i=1}^{k-1} u_{i}^{b_{i}-a_{i}}

1433: $.

1434: Let $F(x,u)$ be a generating function, then $F(x,u)|_{{u}^{b-a}}$

1435: equals the family of coefficients $a_{i}(u)$ at ${u}^{b-a}$ of

1436: $\sum_{i\ge 0}a_i(u)x^i$. We first observe

1437: $$

1438: \Gamma_{n}(a,b)=

1439: \left[1+\sum_{i=1}^{n}(u_{i}+u_{i}^{-1})\right]^n\bigg|_{{u}^{b-a}}

1440: $$

1441: The exponential generating function for $\Gamma_{n}(a,b)$ is

1442: \begin{align*}

1443: \sum_{n \ge 0}\Gamma_{n}(a,b)\frac{x^{n}}{n!}&=\sum_{n \ge 0}

1444: \left[1+\sum_{i=1}^{k-1}(u_{i}+u_{i}^{-1})\right]^n

1445: \bigg|_{{u}^{b-a}}\frac{x^{n}}{n!}\\

1446: &=\sum_{n \ge 0}

1447: \frac{[1+\sum_{i=1}^{k-1}(u_{i}+u_{i}^{-1})]^{n}}{n!}x^{n}

1448: \bigg|_{{u}^{b-a}}\\

1449: &=e^{x}\cdot{\sf exp}[x\sum_{i=1}^{k-1}(u_{i}+u_{i}^{-1})]

1450: \bigg|_{{u}^{b-a}}\\

1451: &=e^{x}\cdot \prod_{i=1}^{k-1}\left({\sf exp}(x(u_{i}+u_{i}^{-1}))

1452: \bigg|_{{u_i}^{b_i-a_i}}\right)\\

1453: \end{align*}

1454: We furthermore derive

1455: \begin{align*}

1456: \sum_{n \ge0}\Gamma_{n}^{+}(a,b)\frac{x^{n}}{n!}& =e^{x}\sum_{\beta \in

1457: B_{k-1}}(-1)^{l(\beta)}\prod_{i=1}^{k-1}

1458: {\sf exp}(x(u_{i}+u_{i}^{-1}))\bigg|_{{u}^{b-\beta(a)}}\\

1459: \end{align*}

1460: and writing $\beta=\epsilon_{h}\,\sigma$ we obtain

1461: \begin{align*}

1462: \sum_{n \ge 0}\Gamma_{n}^{+}(a,b)\frac{x^{n}}{n!}&=

1463: e^{x}\sum_{\sigma \in S_{k-1}}\sum_{h=1}^{k-1}\eta_{h}\,{\rm sgn}

1464: (\sigma)\prod_{i=1}^{k-1}\left({\sf exp}(x(u_{i}+u_{i}^{-1}))

1465: \bigg|_{u_{i}^{b_{i}-\epsilon_{h}a_{\sigma_{i}}}}\right)\\

1466: &=e^{x}\sum_{\sigma \in S_{k-1}}{\rm sgn}(\sigma)\sum_{h=1}^{k-1}

1467: \eta_h\prod_{i=1}^{k-1}\left({\sf exp}(x(u_{i}+u_{i}^{-1}))

1468: \bigg|_{u_{i}^{b_{i}-\epsilon_{h}a_{\sigma_{i}}}}\right)\\

1469: &=e^{x}\sum_{\sigma \in S_{k-1}}{\rm sgn}(\sigma) \left\{

1470: \prod_{i=1}^{k-1}\left({\sf exp}(x(u_{i}+u_{i}^{-1}))

1471: \bigg|_{u_{i}^{b_{i}-a_{\sigma_{i}}}}\right)\right.\\

1472: & \qquad \qquad \qquad \qquad

1473: \left.-\prod_{i=1}^{k-1}\left({\sf exp}(x(u_{i}+u_{i}^{-1}))

1474: \bigg|_{u_{i}^{b_{i}+a_{\sigma_{i}}}}\right)\right\}

1475: \\

1476: &=e^{x}{\sf det}_{{k-1\times k-1}}

1477:             [I_{a_{i}-b_{j}}(2x)-I_{a_{i}+b_{j}}(2x)]|_{i,j=1}^{k-1}\\

1478: \end{align*}

1479: where $\eta_h=\pm 1$ and the lemma follows. $\square$

1480: %%%

1481: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

1482: %%%

1483: 

1484: %%%

1485: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

1486: %%%

1487: {\bf Acknowledgments.}

1488: %%%

1489: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

1490: %%%

1491: We are grateful to Prof.~W.Y.C.~Chen and Prof.~Xin for helpful comments.

1492: Many thanks to J.Z.M.~Gao and F.W.D.~Huang for their help and L.C.~Zuo

1493: for her suggestions.

1494: This work was supported by the 973 Project, the PCSIRT Project of the

1495: Ministry of Education, the Ministry of Science and Technology, and

1496: the National Science Foundation of China.

1497: 

1498: 

1499: \bibliography{bi}

1500: \bibliographystyle{plain}

1501: 

1502: 

1503: 

1504: 

1505: 

1506: 

1507: 

1508: 

1509: %%%

1510: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

1511: %%%

1512: %%%

1513: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

1514: %%%

1515: 

1516: \end{document}

1517: %%%

1518: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

1519: %%%

1520: \begin{center}

1521: \setlength{\unitlength}{2.05pt}

1522: % Drawing generated by LaTeX-CAD 1.9 - requires latexcad.sty

1523: % (c) 1998 John Leis leis@usq.edu.au

1524: \begin{picture}(180,121)

1525: \thinlines \drawpath{-32.0}{-18.0}{106.0}{-18.0}

1526: \drawpath{-24.0}{42.0}{-24.0}{10.0}

1527: \drawcenteredtext{-28.0}{76.0}{\small{1}}

1528: \drawcenteredtext{-28.0}{70.0}{\small{2}}

1529: \drawcenteredtext{-28.0}{64.0}{\small{3}}

1530: \drawcenteredtext{-28.0}{58.0}{\small{4}}

1531: \drawcenteredtext{-28.0}{52.0}{\small{5}}

1532: \drawcenteredtext{-28.0}{46.0}{\small{6}}

1533: \drawcenteredtext{-28.0}{40.0}{\small{7}}

1534: \drawcenteredtext{-28.0}{34.0}{\small{8}}

1535: \drawcenteredtext{-28.0}{28.0}{\small{9}}

1536: \drawcenteredtext{-28.0}{22.0}{\small{10}}

1537: \drawcenteredtext{-28.0}{10.0}{\small{12}}

1538: \drawcenteredtext{-28.0}{16.0}{\small{11}}

1539: \drawpath{-24.0}{10.0}{-24.0}{-4.0}

1540: \drawcenteredtext{-28.0}{4.0}{\small{13}}

1541: \drawcenteredtext{-28.0}{-2.0}{\small{14}}

1542: \drawcenteredtext{-28.0}{-8.0}{\small{15}}

1543: \drawpath{-24.0}{-4.0}{-24.0}{-16.0}

1544: \drawcenteredtext{-28.0}{-14.0}{\small{16}}

1545: \drawcenteredtext{62.0}{88.0}{$\ell$}

1546: \drawpath{-4.0}{86.0}{140.0}{86.0}

1547: \drawpath{-8.0}{80.0}{-8.0}{-16.0}

1548: \drawcenteredtext{8.0}{82.0}{\small{16}}

1549: \drawcenteredtext{18.0}{82.0}{\small{15}}

1550: \drawcenteredtext{28.0}{82.0}{\small{14}}

1551: \drawcenteredtext{38.0}{82.0}{\small{13}}

1552: \drawcenteredtext{48.0}{82.0}{\small{12}}

1553: \drawcenteredtext{58.0}{82.0}{\small{11}}

1554: \drawcenteredtext{68.0}{82.0}{\small{10}}

1555: \drawcenteredtext{78.0}{82.0}{\small{9}}

1556: \drawcenteredtext{88.0}{82.0}{\small{8}}

1557: \drawpath{-32.0}{80.0}{42.0}{80.0}

1558: \drawpath{-24.0}{80.0}{-24.0}{42.0}

1559: \drawcenteredtext{98.0}{82.0}{\small{7}}

1560: \drawpath{-24.0}{92.0}{40.0}{92.0}

1561: \drawcenteredtext{108.0}{82.0}{\small{6}}

1562: \drawcenteredtext{-28.0}{-22.0}{$n$}

1563: \drawcenteredtext{196.0}{86.0}{$n$}

1564: \drawcenteredtext{182.0}{86.0}{Total}

1565: \drawcenteredtext{-16.0}{-22.0}{Total}

1566: \drawpath{-24.0}{92.0}{-32.0}{92.0}

1567: \drawcenteredtext{118.0}{82.0}{\small{5}}

1568: \drawcenteredtext{128.0}{82.0}{\small{4}}

1569: \drawcenteredtext{138.0}{82.0}{\small{3}}

1570: \drawcenteredtext{148.0}{82.0}{\small{2}}

1571: \drawcenteredtext{158.0}{82.0}{\small{1}}

1572: \drawpath{94.0}{92.0}{160.0}{92.0}

1573: \drawpath{140.0}{86.0}{160.0}{86.0}

1574: \drawpath{94.0}{80.0}{160.0}{80.0} \drawpath{40.0}{92.0}{94.0}{92.0}

1575: \drawcenteredtext{8.0}{76.0}{\small{1}}

1576: \drawpath{42.0}{80.0}{94.0}{80.0}

1577: \drawcenteredtext{-16.0}{76.0}{\small{1}}

1578: \drawcenteredtext{-2.0}{58.0}{\small{1}}

1579: \drawcenteredtext{-16.0}{70.0}{\small{1}}

1580: \drawcenteredtext{18.0}{70.0}{\small{1}}

1581: \drawcenteredtext{-2.0}{70.0}{\small{0}}

1582: \drawcenteredtext{8.0}{64.0}{\small{1}}

1583: \drawcenteredtext{28.0}{64.0}{\small{1}}

1584: \drawcenteredtext{-2.0}{46.0}{\small{4}}

1585: \drawcenteredtext{-16.0}{64.0}{\small{2}}

1586: \drawcenteredtext{18.0}{58.0}{\small{3}}

1587: \drawcenteredtext{38.0}{58.0}{\small{1}}

1588: \drawcenteredtext{-2.0}{34.0}{\small{22}}

1589: \drawcenteredtext{8.0}{52.0}{\small{6}}

1590: \drawcenteredtext{28.0}{52.0}{\small{6}}

1591: \drawcenteredtext{48.0}{52.0}{\small{1}}

1592: \drawcenteredtext{-16.0}{22.0}{\small{3334}}

1593: \drawcenteredtext{-16.0}{58.0}{\small{5}}

1594: \drawcenteredtext{18.0}{46.0}{\small{21}}

1595: \drawcenteredtext{38.0}{46.0}{\small{10}}

1596: \drawcenteredtext{58.0}{46.0}{\small{1}}

1597: \drawcenteredtext{-2.0}{22.0}{\small{139}}

1598: \drawcenteredtext{48.0}{40.0}{\small{15}}

1599: \drawcenteredtext{28.0}{40.0}{\small{55}}

1600: \drawcenteredtext{8.0}{40.0}{\small{34}}

1601: \drawcenteredtext{68.0}{40.0}{\small{1}}

1602: \drawcenteredtext{18.0}{22.0}{\small{1317}}

1603: \drawcenteredtext{-16.0}{52.0}{\small{13}}

1604: \drawcenteredtext{18.0}{34.0}{\small{157}}

1605: \drawcenteredtext{38.0}{34.0}{\small{120}}

1606: \drawcenteredtext{58.0}{34.0}{\small{21}}

1607: \drawcenteredtext{78.0}{34.0}{\small{1}}

1608: \drawcenteredtext{38.0}{22.0}{\small{1435}}

1609: \drawcenteredtext{8.0}{28.0}{\small{232}}

1610: \drawcenteredtext{28.0}{28.0}{\small{526}}

1611: \drawcenteredtext{48.0}{28.0}{\small{231}}

1612: \drawcenteredtext{68.0}{28.0}{\small{28}}

1613: \drawcenteredtext{88.0}{28.0}{\small{1}}

1614: \drawcenteredtext{-16.0}{46.0}{\small{36}}

1615: \drawcenteredtext{-16.0}{40.0}{\small{105}}

1616: \drawcenteredtext{-16.0}{34.0}{\small{321}}

1617: \drawcenteredtext{-16.0}{28.0}{\small{1018}}

1618: \drawcenteredtext{58.0}{22.0}{\small{406}}

1619: \drawcenteredtext{78.0}{22.0}{\small{36}}

1620: \drawcenteredtext{98.0}{22.0}{\small{1}}

1621: \drawcenteredtext{-16.0}{16.0}{\small{11216}}

1622: \drawcenteredtext{8.0}{16.0}{\small{1761}}

1623: \drawcenteredtext{28.0}{16.0}{\small{5355}}

1624: \drawcenteredtext{48.0}{16.0}{\small{3388}}

1625: \drawcenteredtext{68.0}{16.0}{\small{666}}

1626: \drawcenteredtext{88.0}{16.0}{\small{45}}

1627: \drawcenteredtext{-16.0}{10.0}{\small{38635}}

1628: \drawcenteredtext{108.0}{16.0}{\small{1}}

1629: \drawcenteredtext{-2.0}{10.0}{\small{979}}

1630: \drawcenteredtext{18.0}{10.0}{\small{11883}}

1631: \drawcenteredtext{38.0}{10.0}{\small{17500}}

1632: \drawcenteredtext{58.0}{10.0}{\small{7182}}

1633: \drawcenteredtext{78.0}{10.0}{\small{1035}}

1634: \drawcenteredtext{98.0}{10.0}{\small{55}}

1635: \drawcenteredtext{118.0}{10.0}{\small{1}}

1636: \drawcenteredtext{-16.0}{4.0}{\small{135835}}

1637: \drawcenteredtext{8.0}{4.0}{\small{14488}}

1638: \drawcenteredtext{28.0}{4.0}{\small{56848}}

1639: \drawcenteredtext{48.0}{4.0}{\small{48888}}

1640: \drawcenteredtext{68.0}{4.0}{\small{14004}}

1641: \drawcenteredtext{88.0}{4.0}{\small{1540}}

1642: \drawcenteredtext{108.0}{4.0}{\small{66}}

1643: \drawcenteredtext{-16.0}{-2.0}{\small{486337}}

1644: \drawcenteredtext{128.0}{4.0}{\small{1}}

1645: \drawcenteredtext{-2.0}{-2.0}{\small{7484}}

1646: \drawcenteredtext{18.0}{-2.0}{\small{113299}}

1647: \drawcenteredtext{38.0}{-2.0}{\small{216468}}

1648: \drawcenteredtext{58.0}{-2.0}{\small{121254}}

1649: \drawcenteredtext{78.0}{-2.0}{\small{25542}}

1650: \drawcenteredtext{98.0}{-2.0}{\small{2211}}

1651: \drawcenteredtext{118.0}{-14.0}{\small{4186}}

1652: \drawcenteredtext{138.0}{-2.0}{\small{1}}

1653: \drawcenteredtext{-16.0}{-8.0}{\small{1769500}}

1654: \drawcenteredtext{8.0}{-8.0}{\small{126748}}

1655: \drawcenteredtext{28.0}{-8.0}{\small{623343}}

1656: \drawcenteredtext{48.0}{-8.0}{\small{698292}}

1657: \drawcenteredtext{68.0}{-8.0}{\small{273834}}

1658: \drawcenteredtext{88.0}{-8.0}{\small{44110}}

1659: \drawcenteredtext{108.0}{-8.0}{\small{3081}}

1660: \drawcenteredtext{128.0}{-8.0}{\small{91}}

1661: \drawcenteredtext{148.0}{-8.0}{\small{1}}

1662: \drawcenteredtext{-16.0}{-14.0}{\small{6531796}}

1663: \drawcenteredtext{-2.0}{-14.0}{\small{61018}}

1664: \drawcenteredtext{18.0}{-14.0}{\small{1127283}}

1665: \drawcenteredtext{38.0}{-14.0}{\small{2709840}}

1666: \drawcenteredtext{58.0}{-14.0}{\small{1983366}}

1667: \drawcenteredtext{78.0}{-14.0}{\small{573210}}

1668: \drawcenteredtext{98.0}{-14.0}{\small{72787}}

1669: \drawcenteredtext{138.0}{-14.0}{\small{105}}

1670: \drawcenteredtext{158.0}{-14.0}{\small{1}}

1671: \drawpath{-24.0}{-16.0}{-24.0}{-18.0}

1672: \drawpath{-8.0}{-16.0}{-8.0}{-18.0}

1673: \drawpath{-32.0}{-26.0}{106.0}{-26.0}

1674: \drawpath{106.0}{-18.0}{162.0}{-18.0}

1675: \drawpath{106.0}{-26.0}{162.0}{-26.0}

1676: \drawpath{160.0}{92.0}{178.0}{92.0}

1677: \drawpath{160.0}{80.0}{178.0}{80.0}

1678: \drawpath{172.0}{-18.0}{172.0}{72.0}

1679: \drawpath{178.0}{92.0}{188.0}{92.0}

1680: \drawpath{178.0}{80.0}{188.0}{80.0}

1681: \drawcenteredtext{-2.0}{-22.0}{\small{0}}

1682: \drawcenteredtext{8.0}{-22.0}{\small{1}}

1683: \drawcenteredtext{18.0}{-22.0}{\small{2}}

1684: \drawcenteredtext{28.0}{-22.0}{\small{3}}

1685: \drawcenteredtext{38.0}{-22.0}{\small{4}}

1686: \drawcenteredtext{48.0}{-22.0}{\small{5}}

1687: \drawcenteredtext{58.0}{-22.0}{\small{6}}

1688: \drawcenteredtext{68.0}{-22.0}{\small{7}}

1689: \drawcenteredtext{78.0}{-22.0}{\small{8}}

1690: \drawcenteredtext{88.0}{-22.0}{\small{9}}

1691: \drawcenteredtext{98.0}{-22.0}{\small{10}}

1692: \drawcenteredtext{108.0}{-22.0}{\small{11}}

1693: \drawcenteredtext{118.0}{-22.0}{\small{12}}

1694: \drawcenteredtext{128.0}{-22.0}{\small{13}}

1695: \drawcenteredtext{138.0}{-22.0}{\small{14}}

1696: \drawcenteredtext{148.0}{-22.0}{\small{15}}

1697: \drawcenteredtext{158.0}{-22.0}{\small{16}}

1698: \drawcenteredtext{118.0}{-2.0}{\small{78}}

1699: \drawpath{162.0}{-18.0}{188.0}{-18.0}

1700: \drawpath{162.0}{-26.0}{188.0}{-26.0}

1701: \drawpath{172.0}{72.0}{172.0}{80.0}

1702: \drawcenteredtext{168.0}{-22.0}{\small{0}}

1703: \drawpath{188.0}{80.0}{190.0}{80.0}

1704: \drawpath{188.0}{92.0}{190.0}{92.0}

1705: \drawpath{190.0}{92.0}{200.0}{92.0}

1706: \drawpath{190.0}{80.0}{200.0}{80.0}

1707: \drawpath{190.0}{80.0}{190.0}{36.0}

1708: \drawpath{190.0}{36.0}{190.0}{-18.0}

1709: \drawpath{190.0}{-18.0}{188.0}{-18.0}

1710: \drawpath{190.0}{-18.0}{198.0}{-18.0}

1711: \drawpath{188.0}{-26.0}{198.0}{-26.0}

1712: \drawpath{198.0}{-18.0}{200.0}{-18.0}

1713: \drawpath{198.0}{-26.0}{200.0}{-26.0}

1714: \drawcenteredtext{28.0}{76.0}{\small{91}}

1715: \drawcenteredtext{48.0}{76.0}{\small{3144}}

1716: \drawcenteredtext{68.0}{76.0}{\small{44724}}

1717: \drawcenteredtext{88.0}{76.0}{\small{269998}}

1718: \drawcenteredtext{108.0}{76.0}{\small{659794}}

1719: \drawcenteredtext{128.0}{76.0}{\small{577123}}

1720: \drawcenteredtext{148.0}{76.0}{\small{136918}}

1721: \drawcenteredtext{166.0}{76.0}{\small{3734}}

1722: \drawcenteredtext{38.0}{70.0}{\small{77}}

1723: \drawcenteredtext{58.0}{70.0}{\small{2263}}

1724: \drawcenteredtext{78.0}{70.0}{\small{25740}}

1725: \drawcenteredtext{98.0}{70.0}{\small{119592}}

1726: \drawcenteredtext{118.0}{70.0}{\small{204037}}

1727: \drawcenteredtext{138.0}{70.0}{\small{107956}}

1728: \drawcenteredtext{158.0}{70.0}{\small{12147}}

1729: \drawpath{160.0}{86.0}{172.0}{86.0}

1730: \drawcenteredtext{48.0}{64.0}{\small{66}}

1731: \drawcenteredtext{68.0}{64.0}{\small{1585}}

1732: \drawcenteredtext{194.0}{76.0}{\small{16}}

1733: \drawcenteredtext{194.0}{70.0}{\small{15}}

1734: \drawcenteredtext{88.0}{64.0}{\small{14362}}

1735: \drawcenteredtext{108.0}{64.0}{\small{49162}}

1736: \drawcenteredtext{128.0}{64.0}{\small{56628}}

1737: \drawcenteredtext{148.0}{64.0}{\small{16729}}

1738: \drawcenteredtext{166.0}{64.0}{\small{538}}

1739: \drawcenteredtext{58.0}{58.0}{\small{55}}

1740: \drawcenteredtext{78.0}{58.0}{\small{1094}}

1741: \drawcenteredtext{98.0}{58.0}{\small{7496}}

1742: \drawcenteredtext{118.0}{58.0}{\small{18241}}

1743: \drawcenteredtext{138.0}{58.0}{\small{12600}}

1744: \drawcenteredtext{158.0}{58.0}{\small{1661}}

1745: \drawcenteredtext{68.0}{52.0}{\small{45}}

1746: \drawcenteredtext{88.0}{52.0}{\small{694}}

1747: \drawcenteredtext{108.0}{52.0}{\small{3543}}

1748: \drawcenteredtext{128.0}{52.0}{\small{5646}}

1749: \drawcenteredtext{148.0}{52.0}{\small{2128}}

1750: \drawcenteredtext{166.0}{52.0}{\small{82}}

1751: \drawcenteredtext{78.0}{46.0}{\small{35}}

1752: \drawcenteredtext{98.0}{46.0}{\small{424}}

1753: \drawcenteredtext{118.0}{46.0}{\small{510}}

1754: \drawcenteredtext{138.0}{46.0}{\small{1462}}

1755: \drawcenteredtext{158.0}{46.0}{\small{238}}

1756: \drawcenteredtext{88.0}{40.0}{\small{28}}

1757: \drawcenteredtext{108.0}{40.0}{\small{246}}

1758: \drawcenteredtext{128.0}{40.0}{\small{571}}

1759: \drawcenteredtext{148.0}{40.0}{\small{286}}

1760: \drawcenteredtext{166.0}{40.0}{\small{13}}

1761: \drawcenteredtext{98.0}{34.0}{\small{21}}

1762: \drawcenteredtext{138.0}{34.0}{\small{176}}

1763: \drawcenteredtext{118.0}{34.0}{\small{130}}

1764: \drawcenteredtext{158.0}{34.0}{\small{36}}

1765: \drawcenteredtext{108.0}{28.0}{\small{15}}

1766: \drawcenteredtext{128.0}{28.0}{\small{61}}

1767: \drawcenteredtext{118.0}{22.0}{\small{10}}

1768: \drawcenteredtext{166.0}{28.0}{\small{2}}

1769: \drawcenteredtext{148.0}{28.0}{\small{40}}

1770: \drawcenteredtext{138.0}{22.0}{\small{24}}

1771: \drawcenteredtext{128.0}{16.0}{\small{6}}

1772: \drawcenteredtext{158.0}{22.0}{\small{5}}

1773: \drawcenteredtext{148.0}{16.0}{\small{7}}

1774: \drawcenteredtext{166.0}{16.0}{\small{0}}

1775: \drawcenteredtext{158.0}{10.0}{\small{1}}

1776: \drawcenteredtext{138.0}{10.0}{\small{3}}

1777: \drawcenteredtext{158.0}{-2.0}{\small{0}}

1778: \drawcenteredtext{166.0}{4.0}{\small{0}}

1779: \drawcenteredtext{148.0}{4.0}{\small{1}}

1780: \drawcenteredtext{166.0}{-8.0}{\small{0}}

1781: \drawcenteredtext{194.0}{64.0}{\small{14}}

1782: \drawcenteredtext{194.0}{58.0}{\small{13}}

1783: \drawcenteredtext{194.0}{52.0}{\small{12}}

1784: \drawcenteredtext{194.0}{46.0}{\small{11}}

1785: \drawcenteredtext{194.0}{40.0}{\small{10}}

1786: \drawcenteredtext{194.0}{34.0}{\small{9}}

1787: \drawcenteredtext{194.0}{28.0}{\small{8}}

1788: \drawcenteredtext{194.0}{22.0}{\small{7}}

1789: \drawcenteredtext{194.0}{16.0}{\small{6}}

1790: \drawcenteredtext{194.0}{10.0}{\small{5}}

1791: \drawcenteredtext{194.0}{4.0}{\small{4}}

1792: \drawcenteredtext{194.0}{-2.0}{\small{3}}

1793: \drawcenteredtext{194.0}{-8.0}{\small{2}}

1794: \drawcenteredtext{194.0}{-14.0}{\small{1}}

1795: \drawcenteredtext{180.0}{76.0}{\small{1695527}}

1796: \drawcenteredtext{180.0}{70.0}{\small{471813}}

1797: \drawcenteredtext{180.0}{64.0}{\small{139071}}

1798: \drawcenteredtext{180.0}{58.0}{\small{41148}}

1799: \drawcenteredtext{180.0}{52.0}{\small{12139}}

1800: \drawcenteredtext{180.0}{40.0}{\small{1145}}

1801: \drawcenteredtext{180.0}{46.0}{\small{2670}}

1802: \drawcenteredtext{180.0}{34.0}{\small{364}}

1803: \drawcenteredtext{180.0}{28.0}{\small{119}}

1804: \drawcenteredtext{180.0}{16.0}{\small{14}}

1805: \drawcenteredtext{180.0}{22.0}{\small{40}}

1806: \drawcenteredtext{180.0}{10.0}{\small{5}}

1807: \drawcenteredtext{180.0}{4.0}{\small{2}}

1808: \drawcenteredtext{180.0}{-2.0}{\small{1}}

1809: \drawcenteredtext{180.0}{-8.0}{\small{1}}

1810: \drawcenteredtext{180.0}{-14.0}{\small{1}}

1811: \drawcenteredtext{166.0}{82.0}{\small{0}}

1812: \drawcenteredtext{70.0}{-34.0}{}

1813: \drawdashline{14.0}{78.0}{124.0}{12.0}

1814: \drawdashline{124.0}{12.0}{152.0}{-6.0}

1815: \drawdashline{152.0}{-6.0}{168.0}{-16.0}

1816: \drawdashline{-2.0}{76.0}{66.0}{36.0}

1817: \drawdashline{148.0}{-14.0}{66.0}{36.0}

1818: \drawdashline{148.0}{-14.0}{152.0}{-16.0}

1819: \end{picture}

1820: \end{center}

1821: 

1822: \par\bigskip

1823: 

1824: \par\bigskip

1825: 

1826: {\small {\bf Table 1.} Lower triangle: for $n=1,\dots, 16$ the numbers of

1827: $3$-noncrossing RNA structures, ${\sf S}_3(n,\ell)$ and ${\sf S}_3(n)$,

1828: respectively. Upper triangle: numbers of restricted $3$-noncrossing RNA

1829: structures, ${\sf S}_3^{(c)}(n,\ell)$ and ${\sf S}_3^{(c)}(n)$.}

1830: 

1831: 

1832: 

1833: \begin{tabular}{c|ccccccccccccccccccccc}

1834: $n$ & \small{1} & \small{2} & \small{3} &\small{4} & \small{5} &

1835: \small{6} & \small{7} & \small{8} & \small{9} & \small{10} &

1836: \small{11} & \small{12} & \small{13} & \small{14} &

1837: \small{15}\\

1838: \hline $S_{3}^{(c)}(n)$ & \small{1} & \small{1} & \small{1}

1839: &\small{4} & \small{11} &\small{31} & \small{92} & \small{285} &

1840: \small{913} & \small{3013} &\small{10180} &\small{35301}

1841: &\small{124853}

1842: &\small{447702} &\small{1626630}\\

1843: \end{tabular}

1844: 

1845: {\small {\bf Table 2.} The first $15$ numbers of $3$-noncrossing circular

1846:         RNA structures.}

1847: