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, eepic,epic}
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:
28: \renewcommand{\baselinestretch}{1.2}% spacing between lines
29:
30: %\hoffset=0truecm
31: %\voffset=0truecm
32: \textwidth=15truecm \textheight=18truecm \baselineskip=0.8truecm
33: \overfullrule=0pt
34: \parskip=0.8\baselineskip
35: \parindent=0truecm
36: \topmargin=0.5truecm \headsep=1.2truecm
37: %\oddsidemargin=0.5in % options for double-side printouts
38: %\evensidemargin=0in
39:
40: %%%
41: %%%
42: %%%%%%%%%%%%%%%%%%%% New Settings %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
43: %%%
44: %%%
45:
46: \theoremstyle{plain}
47: \newtheorem{theorem}{Theorem}
48: \newtheorem{corollary}{Corollary}
49: \newtheorem*{main}{Main~Theorem}
50: \newtheorem{lemma}{Lemma}
51: \newtheorem{proposition}{Proposition}
52: %\newtheorem{remark}{Remark}
53: \theoremstyle{definition}
54: \newtheorem{definition}{Definition}
55:
56: \theoremstyle{example}
57: \newtheorem{example}{Example}
58:
59: \theoremstyle{remark}
60: \newtheorem{remark}{Remark}
61: \numberwithin{equation}{section}
62:
63: % references alphabetically
64: % make relation clearer
65:
66: \begin{document}
67:
68: %%%
69: %%%
70: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
71: %%
72: %%%
73: \title[On $k$-noncrossing partitions]
74: {On $k$-noncrossing partitions}
75: \author{Emma Y. Jin, Jing Qin and Christian M. Reidys$^{\,\star}$}
76: \address{Center for Combinatorics, LPMC-TJKLC \\
77: Nankai University \\
78: Tianjin 300071\\
79: P.R.~China\\
80: Phone: *86-22-2350-6800\\
81: Fax: *86-22-2350-9272}
82: \email{reidys@nankai.edu.cn}
83: \thanks{}
84: \keywords{partition, $k$-noncrossing, $2$-regular, enhanced partition, braid,
85: tangled-diagram,
86: difference equation}
87: \date{October, 2007}
88: \begin{abstract}
89: In this paper we prove a duality between $k$-noncrossing partitions over
90: $[n]=\{1,\dots,n\}$ and $k$-noncrossing braids over $[n-1]$. This duality
91: is derived
92: directly via (generalized) vacillating tableaux which are in correspondence
93: to tangled-diagrams \cite{Reidys:07vac}.
94: We give a combinatorial interpretation of the bijection in terms of the
95: contraction of arcs of tangled-diagrams.
96: Furthermore it induces by restriction a bijection between $k$-noncrossing,
97: $2$-regular partitions over $[n]$ and $k$-noncrossing braids without
98: isolated points over $[n-1]$.
99: Since braids without isolated points correspond to enhanced partitions
100: this allows, using the results of \cite{MIRXIN}, to enumerate $2$-regular,
101: $3$-noncrossing partitions.
102: \end{abstract}
103: \maketitle {{\small
104: %\tableofcontents
105: }}
106:
107:
108:
109: %%%
110: %%%
111: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
112: %%%
113: %%%
114:
115: \section{Introduction and Background}\label{S:1}
116:
117: %%%
118: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
119: %%%
120:
121: In this paper we prove a duality between $k$-noncrossing partitions and
122: braids, a particular type of tangled-diagrams \cite{Reidys:07vac}.
123: The duality implies a bijection between $2$-regular, $k$-noncrossing
124: partitions
125: and $k$-noncrossing braids without isolated points, which are in bijection
126: to enhanced partitions. We then compute the number of $3$-noncrossing,
127: $2$-regular partitions over $[n]=\{1,\dots,n\}$, i.e.~$k$-noncrossing
128: partitions without arcs of the form $(i,i+1)$. The enumeration
129: of $3$-noncrossing, $2$-regular partitions is not entirely trivial. This
130: is due to the fact that the lack of $1$-arcs translates into an asymmetry
131: induced by the nonexistence of the pair of steps ($(\varnothing,+\square_1),
132: (-\square_1,\varnothing)$), where ``$\,\pm\square_i\,$'' denotes the
133: adding/removing of a square in the $i$th row of the shape.
134: We derive the above duality directly via the (generalized) vacillating
135: tableaux \cite{Reidys:07vac} and prove its combinatorial interpretation
136: in terms of the contraction of arcs, originally introduced by
137: Chen~{\it et.al.}~in \cite{Chen-reduction} in the context of a reduction
138: algorithm for noncrossing partitions.
139:
140: Our results imply novel connections between different combinatorial objects
141: and are of conceptual interest.
142: For instance, Bousquet-M\'{e}lou and Xin \cite{MIRXIN} have enumerated
143: $3$-noncrossing partitions and $3$-noncrossing enhanced partitions
144: separately, using kernel methods in nontrivial calculations.
145: By construction enhanced partitions correspond to hesitating tableaux
146: \cite{Chen} which accordingly enumerate braids without isolated points.
147: Our duality theorem implies therefore that either one of these
148: computations would imply the other.
149: Furthermore our results integrate the concepts of vacillating and hesitating
150: tableaux due to Chen~{\it et.al.}~\cite{Chen}.
151: $2$-regular partitions are of particular importance in the context of
152: enumerating RNA tertiary structures with base triples \cite{Reidys:073d}.
153:
154: %%%
155: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
156: %%%
157: \section{Tangled-diagrams and vacillating tableaux}
158: \label{S:2}
159: %%%
160: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
161: %%%
162:
163: In this Section we provide some basics on tangled-diagrams \cite{Reidys:07vac}.
164: A tangled-diagram is a labeled graph, $G_n$, over $[n]$ with degree
165: $\le 2$, represented by drawing its vertices in a horizontal line and
166: its arcs $(i,j)$ in the upper halfplane having the following properties:
167: two arcs $(i_1,j_1)$ and $(i_2,j_2)$ such that $i_1<i_2<j_1<j_2$ are crossing
168: and if $i_1<i_2<j_2<j_1$ they are nesting.
169: Two arcs $(i,j_1)$ and $(i,j_2)$ (common lefthand endpoint) and
170: $j_1<j_2$ can be drawn in two ways: either draw $(i,j_1)$ strictly below
171: $(i,j_2)$ in which case $(i,j_1)$ and $(i,j_2)$ are nesting (at $i$) or
172: draw $(i,j_1)$ starting above $i$ and intersecting $(i,j_2)$ once,
173: in which case $(i,j_1)$ and $(i,j_2)$ are crossing (at $i$):
174: \begin{center}
175: \scalebox{0.25}[0.25]{\includegraphics*[30,710][560,810]{2_new1.eps}}
176: \end{center}
177: and of two arcs $(i,j),(i,j)$, i.e.~where $i$ and $j$ are both:
178: right- and lefthand endpoints are completely analogous. Suppose
179: $i<j<h$ and that we are given two arcs $(i,j)$ and $(j,h)$. Then we
180: can draw them intersecting once or not. In the former case $(i,j)$
181: and $(j,h)$ are called crossing, in the latter noncrossing arcs:
182: \begin{center}
183: \scalebox{0.6}[0.6]{\includegraphics*[60,780][560,830]{yyy5.eps}}
184: \end{center}
185: A $k$-noncrossing braid is a $k$-noncrossing tangled-diagram in
186: which all vertices $j$ of degree two are either incident to loops
187: $(j,j)$ or crossing arcs $(i,j)$ and $(j,h)$, where $i<j<h$. We
188: denote the set of $k$-noncrossing braids over $[n]$ by
189: $\mathcal{B}_k(n)$. For instance
190: \begin{center}
191: \scalebox{0.6}[0.6]{\includegraphics*[60,780][560,830]{yyy06.eps}}
192: \end{center}
193: A shape is a collection of squares, ``$\,\square\,$'', arranged in
194: left-justified rows with weakly decreasing number of squares in each
195: row.
196: %A standard Young tableau is a filling of the squares by numbers
197: % which is strictly decreasing in each row and in each column.
198: % We refer to standard Young tableaux as Young tableaux.
199: A vacillating tableaux $V_{\lambda}^{2n}$ of shape $\lambda$ and
200: length $2n$ is a sequence $(\lambda^{0}, \lambda^{1},\ldots,
201: \lambda^{2n})$ of shapes such that
202: {\sf (i)} $\lambda^{0}=\varnothing$ and $\lambda^{2n}=\lambda,$ and
203: {\sf (ii)} $(\lambda^{2i-1},\lambda^{2i})$ is derived from
204: $\lambda^{2i-2}$, for $1\le i\le n$ by either
205: $(\varnothing,\varnothing)$: do nothing twice;
206: $(-\square,\varnothing)$: first remove a square then do nothing;
207: $(\varnothing,+\square)$: first do nothing then add a square;
208: $(\pm \square,\pm \square)$: add/remove a square at the odd and even steps,
209: respectively. Let $\mathcal{V}_\lambda^{2n}$ denote the set of vacillating
210: tableaux, for instance,
211: \begin{center}
212: \scalebox{0.6}[0.6]{\includegraphics*[10,750][580,830]{yyy_4.eps}}
213: \end{center}
214: We have the following bijection between tangled-diagrams and
215: generalized vacillating tableaux \cite{Reidys:07vac} which
216: integrates the notions of vacillating and hesitating tableaux of
217: Chen~{\it et.al.}~\cite{Chen}. In the following we refer to
218: generalized vacillating tableaux simply as vacillating tableaux.
219: %%%
220: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
221: %%%
222: \begin{theorem}\label{T:bij}
223: There exists a bijection between the set of vacillating tableaux of shape
224: $\varnothing$ and length $2n$, $\mathcal{V}_\varnothing^{2n}$ and the
225: set of tangled-diagrams over $n$ vertices, $\mathcal{G}_n$
226: \begin{equation}
227: \beta\colon \mathcal{V}_{\varnothing}^{2n} \longrightarrow
228: \mathcal{G}_n \ .
229: \end{equation}
230: Furthermore, a tangled-diagram $G_n$ is $k$-noncrossing if and only if all
231: shapes $\lambda^i$
232: in its corresponding vacillating tableau have less than $k$ rows,
233: i.e.~$\beta\colon \mathcal{V}_\varnothing^{2n}\longrightarrow \mathcal{G}_n$
234: maps vacillating tableaux having less than $k$ rows into $k$-noncrossing
235: tangled-diagrams. Furthermore there exists a bijection between the set of
236: $k$-noncrossing and $k$-nonnesting tangled-diagrams.
237: \end{theorem}
238: %%%
239: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
240: %%%
241: Restricting the set of generating step-pairs of vacillating tableaux recovers
242: the bijections of Chen {\it et.al.}~\cite{Chen}:
243: Let $M=\{(\varnothing,\varnothing),(-\square,\varnothing),
244: (\varnothing,+\square)\}$, $\mathcal{V}_{P,k,\varnothing}^{2n}$ and
245: $\mathcal{V}_{B,k,\varnothing}^{2n}$ denote the set of tableaux
246: with less than $k$ rows and generated by
247: $P=M\dot\cup \{(-\square,+\square)\}$ and
248: $B = M \dot\cup \{(+\square,-\square)\}$.
249: Theorem~\ref{T:bij} allows us to identify $\mathcal{V}_{P,k,\varnothing}^{2n}$
250: with $\mathcal{P}_k(n)$ and $\mathcal{V}_{B,k,\varnothing}^{2n}$ with
251: $\mathcal{B}_k(n)$.
252: For partitions and braids we have the following correspondences between the
253: elementary pair-steps and associated tangled-diagram arc-configurations:
254: \begin{center}
255: \scalebox{0.5}[0.5]{\includegraphics*[30,760][560,820]{yyy15.eps}}
256: \end{center}
257: \begin{center}
258: \scalebox{0.5}[0.5]{\includegraphics*[40,760][560,820]{yyy16.eps}}
259: \end{center}
260: %%%
261: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
262: %%%
263:
264: %%%
265: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
266: %%%
267:
268: \section{Main results}\label{S:3}
269:
270: %%%
271: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
272: %%%
273: We now prove the duality between partitions over $[n]$ and braids over
274: $[n-1]$. {\it A posteriori} the above bijection can be proved directly.
275: However, we arrived at this interpretation studying vacillating tableaux of
276: $k$-noncrossing partitions and braids.
277:
278: %%%
279: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
280: %%%
281: \begin{theorem}\label{T:k-noncross}
282: Let $k\in\mathbb{N}$, $k\ge 3$. Then we have the bijection
283: \begin{equation}\label{E:biject1}
284: \vartheta\colon \mathcal{P}_{k}(n)\longrightarrow \mathcal{B}_k(n-1) \ ,
285: \end{equation}
286: where $\vartheta$ has the following property:
287: for any $\pi\in\mathcal{P}_k(n)$ holds:
288: $(i,j)$ is an arc of $\pi$ if and only if $(i,j-1)$ is an arc in
289: $\vartheta(\pi)$.
290: \end{theorem}
291: %%%
292: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
293: %%%
294: \begin{proof}
295: A $k$-noncrossing partition $\pi$ corresponds via Theorem~\ref{T:bij}
296: uniquely to a vacillating tableaux, $V_{\varnothing}^{2n}(\pi)=
297: (\lambda^{i})_{i=0}^{2n}$. Let $\pm \square_h$ denote the adding or
298: subtracting of the rightmost square ``$\,\square\,$'' in the $h$th
299: row in a given shape $\lambda$ and let ``$\,\varnothing\,$'' denote
300: doing nothing.
301: $(\lambda^{i})_{i=0}^{2n}$ uniquely corresponds to a sequence of pairs
302: $\sigma_\pi=((x_i,y_i))_{i=1}^{n}$ where $(x_i,y_i)\in\{(\varnothing,
303: \varnothing), (-\square_j,+\square_h),(\varnothing,+\square_h),
304: (-\square_h,\varnothing)\}$, $1\le h,j\le k-1$ and
305: $x_1=y_{n}=\varnothing$. In the following we shall identify the
306: sequence $(x_i,y_i)_{i=1}^{n}$ with its corresponding sequence of shapes
307: and set
308: \begin{equation}\label{E:varphi}
309: \varphi_1((x_i,y_i)_{i=1}^{n})=(\tilde{x}_{i},\tilde{y}_i)_{i=1}^{n-1} \
310: \quad \text{\rm where}\quad \tilde{x}_i = y_{i}\ \wedge \
311: \tilde{y}_i=x_{i+1} \ .
312: \end{equation}
313: In view of $x_1=y_{n}=\varnothing$ we can conclude that $\varphi_1$ is
314: bijective. Since the vacillating tableaux of a partition is
315: generated by $(-\square,\varnothing)$, $(\varnothing,+\square)$,
316: $(\varnothing,\varnothing)$, $(-\square,+\square)$, we have
317: \begin{equation}\label{E:neu}
318: \forall\, 1\le i\le n-1;\qquad
319: (\tilde{x}_i,\tilde{y}_i)\in
320: \{(\varnothing,\varnothing),\ (+\square_h,\varnothing), \
321: (\varnothing,-\square_h), \ (+\square_h,-\square_j)\} \ ,
322: \end{equation}
323: where $1\le h,j\le k-1$. Let $\varphi_2$ be given by
324: \begin{equation}\label{E:varphi2}
325: \varphi_2((\tilde{x}_i,\tilde{y}_i))=
326: \begin{cases}
327: (\tilde{x}_i,\tilde{y}_i) & \ \text{\rm for } \
328: (\tilde{x}_i,\tilde{y}_i)=(+\square_h,-\square_j) \\
329: (\tilde{y}_i,\tilde{x}_i) & \ \text{\rm otherwise.}
330: \end{cases}
331: \end{equation}
332: $\varphi_2$ has by definition the property
333: $\varphi_2((\tilde{x}_i,\tilde{y}_i))\in \{(-\square_h,\varnothing),
334: (\varnothing,+\square_h),(\varnothing,\varnothing),(+\square_h,-\square_j)\}$.
335: \\
336: {\it Claim $1$.} The mapping
337: $$
338: \vartheta\colon \mathcal{P}_{k}(n)\longrightarrow \mathcal{B}_k(n-1) \ ,\quad
339: \vartheta=\beta\circ \varphi_2\circ\varphi_1
340: \circ\beta^{-1}
341: $$
342: is well-defined and a bijection.\\
343: For arbitrary $\pi\in \mathcal{P}_k(n)$ we set
344: $V_\varnothing^{2n}(\pi)^\dagger=\varphi_2\circ \varphi_1(
345: V_\varnothing^{2n}(\pi))$.
346: By construction $V_\varnothing^{2n}(\pi)^\dagger$ is given by
347: $
348: \varphi_2(\varphi_1(x_i,y_i)_{i=1}^n))=(a_i,b_i)_{i=1}^{n-1}
349: $, where
350: $(a_i,b_i)\in \{(-\square_h,\varnothing),(\varnothing,+\square_h),
351: (\varnothing,\varnothing),(+\square_h,-\square_j)\}$.
352: Its induced sequence of collections of rows of squares
353: $(\mu^i)_{i=0}^{2(n-1)}$ has the following properties:
354: \begin{eqnarray}\label{E:w}
355: && \mu^{2(n-1)}=\lambda^{2n-1}=\varnothing,\\
356: \label{E:right}
357: && \mu^{2j+2}\setminus\mu^{2j+1},\; \mu^{2j+1}\setminus \mu^{2j}\in
358: \{(\varnothing,\varnothing), (\varnothing,+\square), (-\square,\varnothing),
359: (+\square,-\square)\} \\
360: \label{E:shape}
361: && \mu^{2j+1}\neq\lambda^{2j+2}\quad \Longrightarrow\quad
362: \mu^{2j+1}\in \{\lambda^{2j+1},\lambda^{2j+3}\} \ .
363: \end{eqnarray}
364: Eq.~(\ref{E:w}) is obvious and eq.~(\ref{E:right}) follows from
365: eq.~(\ref{E:neu}). By construction of $(\mu^i)_{i=0}^{2(n-1)}$,
366: for $1\le j\le n-1$, $\mu^{2j} = \lambda^{2j+1}$ holds.
367: Suppose $\mu^{2j+1}\neq\lambda^{2j+2}$ for
368: some $0\le j\le n-2$. By definition of $\varphi_2$ only pairs containing
369: ``$\varnothing$'' in at least one coordinate are transposed from which we
370: can conclude $\mu^{2j+1}=\mu^{2j}$ or $\mu^{2j+1}=\mu^{2j+2}$, i.e.
371: $$
372: \diagram
373: & \lambda^{2j+1} \ar@{=}[dl]\rto^{} & \lambda^{2j+2}\ar@{-}[dl]
374: \rto^{} &
375: \lambda^{2j+3} \ar@{=}[dl] \\
376: \mu^{2j} \ar@{=}[r] & \mu^{2j+1} \rto^{} &\mu^{2j+2}
377: \enddiagram
378: \text{\rm or }
379: \diagram
380: & \lambda^{2j+1} \ar@{=}[dl]\rto^{} & \lambda^{2j+2}\ar@{-}[dl]
381: \rto^{} &
382: \lambda^{2j+3} \ar@{=}[dl] \\
383: \mu^{2j} \rto^{} & \mu^{2j+1} \ar@{=}[r] &\mu^{2j+2}
384: \enddiagram
385: $$
386: whence eq.~(\ref{E:shape}). In particular each collection of rows of
387: squares $\mu^i$ is a shape, i.e.~$V_\varnothing^{2n}(\pi)^\dagger$
388: corresponds to a braid.
389: Eq.~(\ref{E:shape}) immediately implies that $(\mu^{i})_{1\le
390: i\le 2(n-1)}$ has at most $k-1$ rows if and only if $(\lambda^{i})_{1
391: \le i\le 2n}$ does. Therefore $\vartheta$ is well-defined. Obviously
392: $\vartheta$ is bijective and Claim $1$ follows.\\
393: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
394: {\it Claim $2$.} For any $\pi\in\mathcal{P}_k(n)$ holds: $(i,j)$ is an arc
395: of $\pi$ if and only if $(i,j-1)$ is an arc in $\vartheta(\pi)$.\\
396: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
397: From the proof of Theorem~\ref{T:bij} \cite{Reidys:07vac} we know that
398: a $\pi$- and $\vartheta(\pi)$-origin at $j$ is equivalent to the
399: existence of a ``$\,+\square\,$'' in the pair-step between the shapes
400: $\lambda^{2j-1}$ and $\lambda^{2j}$ and $\mu^{2j-2}$ and $\mu^{2j}$,
401: respectively. We have the following alternative
402: $$
403: \diagram
404: & \lambda^{2j-1} \ar@{=}[dl]\rto^{+\square} & \lambda^{2j}\ar@{-}[dl]
405: \rto^{-\square} &
406: \lambda^{2j+1} \ar@{=}[dl] \\
407: \mu^{2j-2} \rto^{+\square} & \mu^{2j-1} \rto^{-\square} &\mu^{2j}
408: \enddiagram
409: \quad
410: \diagram
411: & \lambda^{2j-1} \ar@{=}[dl]\rto^{+\square} &
412: \lambda^{2j}\ar@{-}[dl] \rto^{\varnothing} &
413: \lambda^{2j+1} \ar@{=}[dl] \\
414: \mu^{2j-2} \rto^{\varnothing} & \mu^{2j-1} \rto^{+\square} &\mu^{2j}
415: \enddiagram
416: $$
417: It is clear by diagram-chasing that $\pi$ has an origin at $j$ if and
418: only if $\vartheta(\pi)$ does.
419: The situation changes however for endpoints of arcs. A $\pi$- and
420: $\vartheta(\pi)$-endpoint at $j$ is equivalent to a ``$\,-\square\,$'' in the
421: pair-step between $\lambda^{2j-2}$ to $\lambda^{2j-1}$ and $\mu^{2j-2}$
422: to $\mu^{2j}$, respectively. Therefore we have the following two
423: situations:
424: $$
425: \diagram
426: & \lambda^{2j-1} \ar@{=}[dl]\rto^{+\square} & \lambda^{2j}\ar@{-}[dl]
427: \rto^{-\square} &
428: \lambda^{2j+1} \ar@{=}[dl] \\
429: \mu^{2j-2} \rto^{+\square} & \mu^{2j-1} \rto^{-\square} &\mu^{2j}
430: \enddiagram
431: \quad
432: \diagram
433: & \lambda^{2j-1} \ar@{=}[dl]\rto^{\varnothing} &
434: \lambda^{2j}\ar@{-}[dl] \rto^{-\square} &
435: \lambda^{2j+1} \ar@{=}[dl] \\
436: \mu^{2j-2} \rto^{-\square} & \mu^{2j-1} \rto^{\varnothing} &\mu^{2j}
437: \enddiagram
438: $$
439: Again by diagram-chasing we immediately conclude that $j$ is an endpoint in
440: $\vartheta(\pi)$ if and only if $(j+1)$ is an endpoint in $\pi$ and
441: Claim $2$ follows, completing the proof of the theorem.
442: \end{proof}
443: As an illustration of the mapping $\vartheta\colon \mathcal{P}_k(n)
444: \longrightarrow\mathcal{B}_k(n-1)$ we give the following example
445: \begin{center}
446: \scalebox{0.5}[0.5]{\includegraphics*[50,710][620,860]{2_new.eps}}
447: \end{center}
448: Theorem~\ref{T:k-noncross} implies by restriction a bijection
449: between $k$-noncrossing $2$-regular partitions and braids without
450: isolated points, denoted by $\mathcal{B}_k^\dagger(n)$. This is of
451: importance since the former cannot be enumerated via the reflection
452: principle while the latter can: braids without isolated points
453: ``just'' lack the pair-step $(\varnothing,\varnothing)$ which
454: introduces a factor $e^x$ for the generating functions.
455: Consequently, we can enumerate $\mathcal{B}_k^\dagger$ using kernel
456: methods.
457:
458: We will actually give an independent direct proof of this result.
459: For this purpose we interpret $k$-noncrossing braids without isolated
460: points as a subset of $k$-noncrossing partitions. For $\delta\in
461: \mathcal{B}^\dagger_k(n)$ we identify loops with isolated points and
462: crossing arcs $(i,j)$ and $(j,h)$, where $i<j<h$ by noncrossing
463: arcs. We accordingly arrive at the bijection
464: \begin{equation}\label{E:nobraid}
465: f\colon \mathcal{B}_k^\dagger(n)\longrightarrow
466: \{\pi\in \mathcal{P}_k(n)\mid \not \exists\,
467: (i_1,j_1),\dots ,(i_k,j_k);\quad i_1<\dots <i_k=j_1<\dots <j_k \,\} \ .
468: \end{equation}
469: Let $\mathcal{P}_{k,2}(n)$ denote the set of $2$-regular,
470: $k$-noncrossing partitions, i.e.~the set of $k$-noncrossing partitions without
471: arcs of the form $(i,i+1)$.
472: %%%
473: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
474: %%%
475: \begin{theorem}\label{T:bijection-2}
476: Let $k\in\mathbb{N}$, $k\ge 3$. Then we have the bijection
477: \begin{equation}\label{E:biject1}
478: \vartheta\colon \mathcal{P}_{k,2}(n)\longrightarrow
479: \mathcal{B}_k^\dagger(n-1) \ ,
480: \end{equation}
481: where $\vartheta$ is given by Theorem~\ref{T:k-noncross}.
482: \end{theorem}
483: %%%
484: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
485: %%%
486: \begin{proof}
487: By construction, $\vartheta$ maps tangled-diagrams over $[n]$
488: into tangled diagrams over $[n-1]$.
489: Since there exist no arcs of the form $(i,i+1)$, $\vartheta(\pi)$ is,
490: for any $\pi\in \mathcal{P}_{k,2}(n)$ loop-free. By construction,
491: $\vartheta$ preserves the orientation of arcs, whence $\vartheta(\pi)$
492: is a partition.\\
493: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
494: {\it Claim.} $\vartheta\colon \mathcal{P}_{k,2}(n)\longrightarrow
495: \mathcal{B}_k^\dagger(n-1)$ is well-defined.\\
496: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
497: We first prove that $\vartheta(\pi)$ is $k$-noncrossing.
498: Suppose there exist $k$ mutually crossing arcs, $(i_s,j_s)$, $s=1,\dots ,k$
499: in $\vartheta(\pi)$. Since $\vartheta(\pi)$ is a partition we have
500: $i_1<\dots <i_k<j_1<\dots <j_k$. Accordingly, we obtain for the partition
501: $\pi \in \mathcal{P}_{k,2}(n)$ the $k$ arcs
502: $(i_s,j_s+1)$, $s=1,\dots ,k$ where $i_1<\dots <i_k<j_1+1<\dots <j_k+1$,
503: which is impossible since $\pi$ is $k$-noncrossing.
504: We next show that $\vartheta(\pi)$ is a $k$-noncrossing braid. If
505: $\vartheta(\pi)$ is not a $k$-noncrossing braid, then according to
506: eq.~(\ref{E:nobraid}) $\vartheta(\pi)$
507: contains $k$ arcs of the form $(i_1,j_1),\dots (i_k,j_k)$ such that
508: $i_1< \dots <i_k=j_1<\dots <j_k$ holds. Then $\pi$ contains the arcs
509: $(i_1,j_1+1)$, $(i_k,j_k+1)$ where $i_1<\dots <i_k<j_1+1<\dots <j_k+1$,
510: which is impossible since these arcs are a set of $k$ mutually crossing arcs
511: and the claim follows.\\
512: %%%%%%%%%%%%%% inverse %%%%%%%%%%%%%%%%%%
513: {\it Claim.} $\vartheta$ is bijective.\\
514: Clearly $\vartheta$ is injective and it remains to prove surjectivity.
515: For any $k$-noncrossing braid $\delta$ there exists some $2$-regular
516: partition $\pi$ such that $\vartheta(\pi)=\delta$.
517: We have to show that $\pi$ is $k$-noncrossing.
518: Let $M'=\{(i_1,j_1),\dots ,(i_k,j_k)\}$ be a set of $k$ mutually
519: crossing arcs, i.e.~$i_1<\dots <i_k<j_1<\dots <j_k$. Then we have in
520: $\vartheta(\pi)$ the arcs $(i_s,j_s-1)$,
521: $s=1,\dots ,k$ and $i_1<\dots <i_k\le
522: j_1-1<\dots <j_k-1$. If $M=\{(i_1,j_1-1),\dots ,(i_k,j_k-1)\}$ is
523: $k$-noncrossing then we can conclude $i_k=j_1-1$. Therefore
524: $M=\{(i_1,j_1-1),\dots ,(i_k,j_k-1)\}$, where $i_k=j_1-1$ which is,
525: in view of eq.~(\ref{E:nobraid}) impossible in $k$-noncrossing braids.
526: By transposition we have thus proved that any $\vartheta$-preimage
527: is necessarily a $k$-noncrossing partition, whence the claim and the proof
528: of the theorem is complete.
529: \end{proof}
530:
531: Theorem~\ref{T:bijection-2} allows for lattice path enumeration of
532: $\mathcal{P}_{k,2}(n)$. The main difficulty lies the kernel-computation
533: \cite{Mohanty79} and at present time there exists no such formula for
534: $k>3$. However, for $\mathcal{B}_3^\dagger(n-1)$ we have in the following
535: enumerative result.
536: %%%
537: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
538: %%%
539: \begin{theorem}\cite{MIRXIN,Reidys:073d}\label{T:braid}
540: The number of $3$-noncrossing braids without isolated points over $[n]$,
541: $\rho_3(n)$, is given by
542: \begin{eqnarray*}{\label{E:braidnum1}}
543: \rho_3(n) &=& \sum_{s\in\mathbb{Z}}\left[\beta_{n}(1,0,s)-\beta_{n}(1,-1,s)-
544: \beta_{n}(1,-4,s)+\beta_{n}(1,-3,s)\right.\\
545: & & -\beta_{n}(3,4,s)+\beta_{n}(3,3,s)+\beta_n(3,0,s)-\beta_{n}(3,1,s)\\
546: & & \left.
547: +\beta_{n}(2,5,s)-\beta_{n}(2,4,s)-\beta_{n}(2,1,s)+\beta_{n}(2,2,s))\right]
548: \ ,
549: \end{eqnarray*}
550: where $\beta_{n}(t,m,s)=\frac{t}{n+1}{n+1\choose s}{n+1\choose t+s}
551: {n+1\choose s+m}$. Furthermore $\rho_3(n)$ satisfies the recursion
552: \begin{equation}\label{E:recursion-b}
553: \alpha_1(n)\, \rho_{3}(n)+
554: \alpha_2(n)\, \rho_3(n+1)+
555: \alpha_3(n)\, \rho_{3}(n+2)-
556: \alpha_4(n)\, \rho_{3}(n+3)=0 \ ,
557: \end{equation}
558: where $\alpha_1(n) = 8(n+2)(n+3)(n+1)$, $\alpha_2(n) =
559: 3(n+2)(5n^2+47n+104)$, $\alpha_3(n) = 3(n+4)(2n+11)(n+7)$ and
560: $\alpha_4(n)= (n+9)(n+8)(n+7)$ and
561: \begin{equation}
562: \rho_{3}(n)\sim K \ 8^{n}n^{-7}(1+c_{1}/n+c_{2}/n^2+c_3/n^3),
563: \end{equation}
564: where $K=6686.408973$, $c_1=-28,\
565: c_2=455.77778$ and $c_3=-5651.160494$.
566: \end{theorem}
567: %%%
568: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
569: %%%
570: The theorem has two parts: the first is the exact formula resulting from the
571: kernel computation \cite{MIRXIN} and the second is the asymptotic formula
572: \cite{Reidys:073d}.
573: In \cite{MIRXIN} the exact formula is computed, the authors also prove an
574: asymptotic formula.
575: In \cite{Reidys:073d} an improved asymptotic formula is given which is based
576: the analytic theory of singular difference equations developed by Birkhoff
577: and Trjitzinsky \cite{Birkhoff,T:wimp}. To keep the paper self-contained we
578: prove Theorem~\ref{T:braid} in the Section~\ref{S:braid}.
579:
580: \begin{remark}
581: The enumeration results for $\mathcal{B}_3^\dagger(n)$ summarized in
582: Theorem~\ref{T:braid} imply trivially the enumeration of $\mathcal{B}_3(n)$.
583: According to the duality between braids and partitions we have therefore
584: obtained the enumeration of $3$-noncrossing partitions.
585: \end{remark}
586: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
587:
588:
589:
590: %%%
591: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
592: %%%
593:
594: \section{Proof of Theorem~\ref{T:braid}}\label{S:braid}
595:
596:
597: %%%
598: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
599: %%%
600:
601: We have $k=3$, i.e.~walks induced by the vacillating braid-tableaux
602: in $\mathbb{Z}^2$, starting and ending at $(1,0)$. Via the
603: reflection principle we reduce the enumeration of these walks which
604: remain in the first quadrant and never touch the diagonal $x=y$
605: to the enumeration of lattice walks in the first quadrant
606: starting and ending at $(1,0)$ and starting at $(1,0)$ and ending at $(0,1)$,
607: respectively. Let $h(i,j,l)$ be the number of walks of length $l$ that end
608: at $(i,j)$ and let $H(x,y;t) = \sum_{i,j,l}h(i,j,l)x^iy^jt^{l}$.
609: We set $\bar{x}=x^{-1}$.\\
610: %%%
611: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
612: %%%
613: {\bf Claim $1$.}
614: The series $H(x,0;t)$ and $H(0,\bar{x};t)$ satisfy
615: \begin{align}\label{E:11}
616: t^2x(x+1)H(x,0;t)&=PT_{x}(x^2Y_0-\bar{x}^2Y_{0}^3+\bar{x}^3Y_{0}^2)\\
617: \label{E:22}
618: t^2\bar{x}(\bar{x}+1)H(0,\bar{x};t)&=NT_{x}(x^2Y_0-\bar{x}^2Y_{0}^3+
619: \bar{x}^3Y_{0}^2),
620: \end{align}
621: where the operator $PT_{x}$($NT_{x}$) extracts positive(negative)
622: powers of $x$ in series of $\mathbb{Q}[x,\bar{x}][[t]]$.\\
623: %%%
624: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
625: %%%
626: To prove the Claim $1$ we observe that the kernel of
627: \begin{eqnarray*}
628: H(x,y;t)-x & = &
629: (x+y+\bar{x}+\bar{y}+x\bar{y}+y\bar{x}+y\bar{y}+x\bar{x})\,t^2 H(x,y;t) \\
630: && -t^2\, (x\bar{y}+\bar{y})\, H(x,0;t)-t^2\, (y\bar{x}+\bar{x})\,H(0,y;t)
631: \end{eqnarray*}
632: is given by:
633: \begin{equation}{\label{E:braidkernel}}
634: K_{\mathcal{B}_3}(x,y;t)=xy-t^2(x^2y+xy^2+y+x+x^2+y^2+2xy) \ .
635: \end{equation}
636: $K_{\mathcal{B}_3}(x,y;t)$ is an irreducible polynomial of degree $2$
637: over $\mathbb{Q}(y,t)$ having the two roots $Y_0=Y_0(x,t)$ and
638: $Y_1=Y_1(x,t)$. Only $Y_0$ is a power series with positive coefficients
639: in $t^2$:
640: \begin{equation}{\label{E:Y_0}}
641: Y_0 = \frac{1-t^2(x+2+\bar{x})-
642: \sqrt{(1-t^2(x+2+\bar{x}))^2-4t^4x\,(1+\bar{x})^2}}
643: {2t^2(\bar{x}+1)}
644: \end{equation}
645: i.e.~$Y_0(x,t)=(1+x)t^2+(x(x+1)(\bar{x}+1)^2)t^4+O(t^6)$.
646: Furthermore we have $Y_0\,Y_1=x$ and
647: \begin{equation}{\label{E:braidkernel2}}
648: x^2\,\bar{y} \,K_{\mathcal{B}_3}(\bar{x}y,y;t)= K_{\mathcal{B}_3}(x,y;t) =
649: x^3\,K_{\mathcal{B}_3}(\bar{x}y,\bar{x};t) \ .
650: \end{equation}
651: Eq.~(\ref{E:braidkernel2}) implies
652: $K_{\mathcal{B}_3}(\bar{x}Y_0,\bar{x};t)=
653: K_{\mathcal{B}_3}(\bar{x}Y_0,Y_0;t)=K_{\mathcal{B}_3}(x,Y_0;t)=0$ and
654: we accordingly obtain
655: \begin{align}
656: x^2Y_0&=t^2x(x+1)H(x,0;t)+t^2Y_0(Y_0+1)H(0,Y_0;t)\\
657: \bar{x}^2Y_0^3&=t^2\bar{x}Y_0(\bar{x}Y_0+1)H(\bar{x}Y_0,0;t)+
658: t^2Y_0(Y_0+1)H(0,Y_0;t)\\
659: \bar{x}^3Y_{0}^2&=t^2\bar{x}Y_{0}(\bar{x}Y_{0}+1)H(\bar{x}Y_{0},0;t)+
660: t^2\bar{x}(\bar{x}+1)H(0,\bar{x};t) \ .
661: \end{align}
662: We next eliminate the terms $H(0,Y_{0};t)$ and $H(\bar{x}Y_{0},0;t)$
663: and arrive at
664: \begin{equation}{\label{E:braideqn}}
665: x^{2}Y_{0}-\bar{x}^2Y_{0}^3+\bar{x}^3Y_{0}^2=
666: t^2x(x+1)H(x,0;t)+t^2\bar{x}(\bar{x}+1)H(0,\bar{x};t) \ .
667: \end{equation}
668: Since $t^2x(x+1)H(x,0;t)$ and $t^2\bar{x}H(0,\bar{x};t)$ have only
669: positive and negative powers of $x$, respectively, we can conclude
670: \begin{eqnarray*}
671: t^2x(x+1)H(x,0;t) &=& PT_{x}(x^2Y_0-\bar{x}^2Y_{0}^3+\bar{x}^3Y_{0}^2)\\
672: t^2\bar{x}(\bar{x}+1)H(0,\bar{x};t) &=& NT_{x}(x^2Y_0-
673: \bar{x}^2Y_{0}^3+\bar{x}^3Y_{0}^2) \ .
674: \end{eqnarray*}
675: {\bf Claim $2$.}
676: Let $CT_x$ denote the constant coefficient of a Laurent-series
677: $\sum_{i\in I}a_ix^i$. Then we have
678: \begin{equation}\label{E:claim1}
679: \rho_3(n)= [t^{2n+2}]CT_{x}((1-x-x^4+x^3)Y_{0}+
680: (-\bar{x}^{4}+\bar{x}^{3}+1-\bar{x})Y_{0}^{3}+(\bar{x}^{5}-
681: \bar{x}^{4}-\bar{x}+\bar{x}^2)Y_{0}^2)\ .
682: \end{equation}
683: To prove Claim $2$ we write $\rho_3(n) =[xt^{2n}]H(x,0;t)-[yt^{2n}]H(0,y;t)$
684: and interpret the terms $[xt^{2n}]H(x,0;t)$ and $[yt^{2n}]H(0,y;t)$
685: via eq.~(\ref{E:11}) and eq.~(\ref{E:22}):
686: \begin{align*}
687: [xt^{2n}]H(x,0;t) &=
688: [x^2t^{2n+2}]PT_{x}(x^2Y_{0}-\bar{x}^2Y_{0}^3+\bar{x}^3Y_{0}^2)
689: -[xt^{2n+2}]PT_{x}(x^2Y_{0}-\bar{x}^2Y_{0}^3+\bar{x}^3Y_{0}^2) \\
690: [yt^{2n}]H(0,y;t) &= [\bar{x}t^{2n}]H(0,\bar{x};t)\\
691: &=[\bar{x}^2t^{2n+2}]\, NT_{x}(x^2Y_{0}-\bar{x}^2Y_{0}^3+\bar{x}^3Y_{0}^2)
692: -[\bar{x}t^{2n+2}]\, NT_{x}(x^2Y_{0}-\bar{x}^2Y_{0}^3+\bar{x}^3Y_{0}^2) \ .
693: \end{align*}
694: We can combine these equations and obtain
695: \begin{align*}
696: \rho_3(n) &= [t^{2n+2}]CT_{x}(\bar{x}^{2}-\bar{x}-x^2+x)(x^2Y_{0}-
697: \bar{x}^2Y_{0}^{3}+\bar{x}^{3}Y_{0}^{2})\\
698: &{\label{E:12term}}=[t^{2n+2}]CT_{x}((1-x-x^4+x^3)Y_{0}+
699: (-\bar{x}^{4}+\bar{x}^{3}+1-\bar{x})Y_{0}^{3}+(\bar{x}^{5}-
700: \bar{x}^{4}-\bar{x}+\bar{x}^2)Y_{0}^2) \ .
701: \end{align*}
702: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
703: {\bf Claim $3$.} Suppose $Y_0$ is the solution of $K_{\mathcal{B}_3}(x,y;t)=0$
704: with positive coefficients in $t^2$ of eq.~(\ref{E:Y_0}). Then we have
705: \begin{equation}
706: [x^mt^{2n+2}]Y_{0}^{k}=\frac{k}{n+1}{n+1\choose
707: s}{n+1\choose k+s}{n+1\choose s+m} \ .
708: \end{equation}
709: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
710: Since $K_{\mathcal{B}_3}(x,Y_0;t)=0$, (eq.~(\ref{E:braidkernel})) we have
711: $Y_{0}=t^{2}(\bar{x}+1)(x+Y_{0})(1+Y_{0}).$ Let
712: $\mathcal{G}(t^2)=(\bar{x}+1)(x+t^2)(1+t^2)$. We derive
713: \begin{align*}
714: [t^{2n+2}]Y_{0}^{k}&=
715: \frac{k}{n+1}[t^{2(n+1-k)}](\bar{x}+1)^{n+1}(x+t^2)^{n+1}(1+t^2)^{n+1}\\
716: &=\frac{k}{n+1}\left(\sum_{s=0}^{n+1-k}(\bar{x}+1)^{n+1}{n+1\choose s}
717: {n+s\choose n+1-s-k}x^{n+1-s}\right).
718: \end{align*}
719: We can conclude from this
720: \begin{equation}
721: [x^{m}t^{2n+2}]Y_{0}^{k}=\frac{k}{n+1}\sum_{s=0}^{n+1}{n+1\choose
722: s}{n+1\choose k+s}{n+1\choose s+m}.
723: \end{equation}
724: and Claim $3$ follows.
725: In order to prove the first assertion of the theorem, we calculate the
726: first term $[t^{2n+2}]CT_{x}((1-x-x^4+x^3)Y_{0}$ of eq.~(\ref{E:claim1}).
727: The terms
728: $(-\bar{x}^{4}+\bar{x}^{3}+1-\bar{x})Y_{0}^{3}$ and $(\bar{x}^{5}-
729: \bar{x}^{4}-\bar{x}+\bar{x}^2)Y_{0}^2$
730: can be computed analogously:
731: \begin{align*}
732: [t^{2n+2}]CT_{x}((1-x-x^4+x^3)Y_{0}&=[x^{0}t^{2n+2}]Y_{0}-
733: [x^{-1}t^{2n+2}]Y_{0}-[x^{-4}t^{2n+2}]Y_{0}+[x^{-3}t^{2n+2}]Y_0\\
734: &=\sum_{s=0}^{n+1}(\beta_{n}(1,0,s)-\beta_{n}(1,-1,s)-
735: \beta_{n}(1,-4,s)+\beta_{n}(1,-3,s))\ ,
736: \end{align*}
737: where $\beta_{n}(k,m,s)=\frac{k}{n+1}{n+1\choose s}{n+1\choose
738: k+s}{n+1\choose s+m}$. Using eq.~(\ref{E:braidnum1}) the recursion
739: follows from Zeilberger's algorithms {\cite{Wilf}} using MAPLE.\\
740: {\bf Claim $4$.}
741: There exist some $K>0$ and $c_{1},c_{2},c_{3}\dots$ such that
742: \begin{equation}
743: \rho_{3}(n)\sim K \ 8^{n}n^{-7}(1+c_{1}/n+c_{2}/n^2+c_3/n^3\cdots).
744: \end{equation}
745: The theory of singular difference equations \cite{Birkhoff} guarantees
746: the existence of $3$ linearly independent formal series solutions (FSS) for
747: eq.~(\ref{E:recursion-b}). We set
748: \begin{equation}
749: \rho_{3}(n) = E(n)K(n) \quad E(n)=e^{\mu_0n\ln n+\mu_{1}n}n^{\theta}
750: \end{equation}
751: where $K(n)=\exp\{\alpha_{1}n^{\beta+\alpha_2n^{\beta-1/\rho+\cdots}}\}$,
752: $\alpha_1\neq 0$, $\beta=j/\rho$, and $0\leq j<\rho$.
753: We immediately derive setting $\lambda =e^{\mu_{0}+\mu_{1}}$
754: \begin{align*}
755: \frac{\rho_{3}(n+k)}{\rho_{3}(n)}&=n^{\mu_{0}k}\lambda^{k}
756: \{1+\frac{k\theta+k^2\mu_{0}/2}{n}+\cdots\} \\
757: & \quad \ \exp\{\alpha_1\beta kn^{\beta-1}+\alpha_2(\beta-\frac{1}{\rho})
758: kn^{\beta-1/\rho-1+\cdots}\},
759: \end{align*}
760: and arrive at
761: \begin{align*}
762: 0=1+&\frac{15}{8}\{1+\frac{\theta+\mu_0/2+\frac{27}{5}}{n}+
763: \cdots\}\xi\{1+(\alpha_1\beta
764: n^{\beta-1}+\alpha_2(\beta-1/\rho)n^{\beta-1/\rho-1}+\cdots)+\cdots\}\\
765: +&\frac{3}{4}\{1+\frac{2\theta+2\mu_0+\frac{21}{2}}{n}+\cdots\}
766: \xi^2\{1+(2\alpha_1\beta
767: n^{\beta-1}+2\alpha_2(\beta-1/\rho)n^{\beta-1/\rho-1}+\cdots)+\cdots\}\\
768: -&\frac{1}{8}\{1+\frac{3\theta+9\mu_0/2+18}{n}+\cdots\}
769: \xi^3\{1+(3\alpha_1\beta
770: n^{\beta-1}+3\alpha_2(\beta-1/\rho)n^{\beta-1/\rho-1}+\cdots)+\cdots\}.
771: \end{align*}
772: First we consider the maximum power of $n$, which is zero. In view
773: of $1=\frac{1}{8}n^{3\mu_0}\lambda^3$ we obtain $\mu_0=0$. This
774: implies $\rho=1$ since $\rho\geq 1$ and $\rho$ should be the
775: smallest integer s.t. $\rho\mu_0\in \mathbb{N}$. Equating the
776: constant terms again, we obtain that $\lambda$ is indeed a root of
777: the cubic polynomial $P(X)$
778: $
779: P(X) = 1+\frac{15}{8}X+\frac{3}{4}X^2-\frac{1}{8}X^3
780: $.
781: Therefore we have $\lambda=8$ or $-1$. Notice that $0\leq \beta <1$
782: implies $\beta=0$. Otherwise, equating the coefficient of
783: $n^{\beta-1}$ implies $\alpha_1=0$, which is impossible. It remains
784: to compute $\theta$. For this purpose we equate the coefficient of
785: $n^{-1}$, i.e.~
786: $
787: 8\frac{15}{8}(\theta+\frac{27}{5})+8^2\frac{3}{4}
788: (\frac{21}{2}+2\theta)-8^3\frac{1}{8}(18+3\theta)=0
789: $
790: from which we can conclude $\theta=-7$. Since $\rho_3(n)$ is
791: monotone increasing $\rho_3(n)$ coincides with the only
792: monotonously increasing FSS, given by
793: \begin{equation}
794: \rho_{3}(n)\sim K \cdot 8^{n}\cdot
795: n^{-7}(1+c_{1}/n+c_{2}/n^2+c_3/n^3\cdots)
796: \end{equation}
797: for some $K>0$ and constants $c_1,c_2,c_3$ and the proof of the
798: Claim $4$ is complete. \\
799: Equating the coefficients of $n^{-2},\ n^{-3}$ and $n^{-4}$, ($2268+81c_1=0$,
800: $1683c_1+162c_2-26712=0$ and $-32547c_1+729c_2+129654+243c_3=0$)
801: we obtain $c_1=-28$, $c_2=455.778$ and $c_3=-5651.160494$ and finally we
802: compute $K=6686.408973$ numerically, completing the proof of
803: Theorem~\ref{T:braid}.
804:
805: %%%%%%
806: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
807: %%%
808: {\bf Acknowledgments.}
809: %%%
810: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
811: %%%
812: We are grateful to Prof.~W.Y.C.~Chen for stimulating discussions. This work
813: was supported by the 973 Project, the PCSIRT Project of the Ministry
814: of Education, the Ministry of Science and Technology, and the
815: National Science Foundation of China.
816:
817:
818: \bibliographystyle{amsplain}
819: \begin{thebibliography}{10}
820:
821: \bibitem{MIRXIN} Mireille Bousquet-M\'{e}lou and Guoce Xin,
822: \textit{On partitions avoiding 3-crossings.} S\'{e}minaire
823: Lotharingien de Combinatoire, \textbf{54} (2006), Article B54c.
824:
825: \bibitem{Birkhoff} George D. Birkhoff, \textit{Formal theory of irregular difference equations} Acta Math. \textbf{54}
826: (1930), 205-246.
827:
828: \bibitem{T:wimp} George D. Birkhoff and W. J. Trjitzinsky, \textit{Analytic theory of singular difference equations} Acta Math., \textbf{60}
829: (1932), 1--89.
830:
831: \bibitem{Chen} William Y. C. Chen, Eva Y.P. Deng, Rosena R.X. Du,
832: Richard P. Stanley and Catherine H. Yan, \textit{Crossings and
833: Nestings of Matchings and Partitions.} Trans. Amer. Math. Soc.
834: \textbf{359} (2007), No. 4, 1555--1575.
835:
836: \bibitem{Chen-reduction}
837: W.Y.C. Chen, E.Y.P. Deng, R.R.X. Du, \textit{Reduction of m-regular
838: noncrossing partitions}, Europ. J. Combin. \textbf{26} (2005), No. 2,
839: 237-243.
840:
841: \bibitem{Reidys:07vac} William Y. C. Chen, Jing Qin and Christian M. Reidys,
842: \textit{Crossings and Nestings of tangled-diagrams.} Submitted.
843:
844: \bibitem{Mohanty79} S. G. Mohanty, \textit{Lattice Path Counting and
845: Applications}, Academic Press, New York, 1979.
846:
847: \bibitem{Wilf} M. Petov\v{s}ek, H.S. Wilf and D. Zeiberger,
848: \textit{A=B}. A K Peter Ltd., Wellesey, MA, 1996.
849:
850: \bibitem{Reidys:073d} Jing Qin and Christian M. Reidys,
851: \textit{A combinatorial framework for RNA-tertiary interactions.} Submitted.
852:
853: \bibitem{wimp} Jet Wimp and Doron Zeilberger, \textit{Resurrecting the
854: asymptotics of linear recurrences}, Journal of Mathmatical analysis and
855: applications, \textbf{3} (1985), 162--176.
856:
857: \end{thebibliography}
858:
859: \end{document}
860: