1: \documentclass[12pt]{article}

2: 

3: \usepackage{a4wide}

4: \usepackage{amssymb,amsmath,amsthm}

5: 

6: \usepackage{rotating}

7: \usepackage{graphicx}

8: \usepackage{latexsym}

9: \usepackage{bm}

10: \usepackage{mathrsfs}

11: 

12: 

13: \newcommand{\AGf}{\ensuremath{\mathcal{A}}}

14: \newcommand{\PGf}{\ensuremath{\mathcal{P}}}

15: \newcommand{\QGf}{\ensuremath{\mathcal{Q}}}

16: \newcommand{\PPR}{\ensuremath{\mathcal{P}_{\rm Rot}}}

17: \newcommand{\PP}[1]{\ensuremath{\mathcal{P}^{(#1)}}}

18: \newcommand{\PPM}[2]{\ensuremath{\mathcal{P}^{(#1)}_{#2}}}

19: \newcommand{\StM}[1]{\ensuremath{\mathcal{P}}^{\,\Box\,(#1)}}

20: \newcommand{\StMR}[1]{\ensuremath{\mathcal{P}}_{\,#1}^{\,\Box,(reg)}}

21: \newcommand{\StMM}[2]{\ensuremath{\mathcal{P}}_{#2}^{\,\Box\,(#1)}}

22: \newcommand{\Pol}{\ensuremath{\mathscr{P}}}

23: \newcommand{\Sign}[2]{\ensuremath{{ \rm Sign}\left  ( [x^{#2}] #1 \right )}}

24: \newcommand{\La}[2]{\ensuremath{A^{(#1)}_{#2}}}

25: \newcommand{\Lb}[2]{\ensuremath{B^{(#1)}_{#2}}}

26: \newcommand{\Lc}[2]{\ensuremath{C^{(#1)}_{#2}}}

27: \newcommand{\ave}[1]{\langle #1 \rangle}

28: 

29: \newcommand{\FM}[1]{\ensuremath{{\cal F}}^{\,\Box\,(#1)}}

30: 

31: %\theoremstyle{definition}

32: 

33: \newtheorem{theorem}{Theorem}

34: \newtheorem{fact}{Fact}

35: \newtheorem{prop}{Proposition}

36: \newtheorem{lem}{Lemma}

37: \newtheorem{definition}{Definition}

38: \newtheorem{conjecture}{Conjecture}

39: 

40: 

41: \title{Area distribution and scaling function \\for punctured polygons}

42: 

43: \author{\sc Christoph Richard\dag, Iwan Jensen\ddag{} and Anthony J. Guttmann\ddag\\

44: \normalsize

45: \dag Fakult\"at f\"ur Mathematik, Universit\"at Bielefeld,\\

46: \normalsize

47: Postfach 10 01 31, 33501 Bielefeld, Germany\\

48: \normalsize

49: \ddag ARC Centre of Excellence for Mathematics and Statistics of Complex Systems,\\ 

50: \normalsize

51: Department of Mathematics and Statistics, 

52: The University of Melbourne,\\ 

53: \normalsize

54: Victoria 3010, Australia}

55: 

56: \begin{document}

57: 

58: \maketitle

59: 

60: \centerline{Mathematics Subject Classifications: 05A15, 05A16}

61: \centerline{PACS numbers: 05.50.+q, 05.70.Jk, 02.10.Ox}

62: 

63: \begin{abstract}

64: Punctured polygons are polygons with internal holes which are 

65: also polygons. The external and internal polygons are of the 

66: same type, and they are mutually as well as self-avoiding. 

67: Based on an assumption about the limiting area distribution 

68: for unpunctured polygons, we rigorously analyse the effect of a 

69: finite number of punctures on the limiting area distribution in 

70: a uniform ensemble, where punctured polygons with equal perimeter 

71: have the same probability of occurrence. Our analysis 

72: leads to conjectures about the scaling behaviour 

73: of the models.

74: 

75: We also analyse exact enumeration data. For staircase polygons 

76: with punctures of fixed size, this yields explicit expressions

77: for the generating functions of the first few area moments.

78: For staircase polygons with punctures of arbitrary size,

79: a careful numerical analysis yields very accurate estimates 

80: for the area moments.

81: Interestingly, we find that the leading correction term for each area moment 

82: is proportional to the corresponding area moment with one less puncture. 

83: We finally analyse corresponding quantities for punctured self-avoiding 

84: polygons and find agreement with the conjectured formulas to at 

85: least 3--4 significant digits.

86: \end{abstract}

87: 

88: \maketitle

89: 

90: \section{Introduction}

91: 

92: The behaviour of planar self-avoiding walks (SAW) and polygons (SAP) is one 

93: of the classical unsolved problems, not only of algebraic combinatorics, 

94: but also of chemistry and of physics \cite{MSbook,HughesV1,Vanderzande}.

95: In the field of algebraic combinatorics, it is a classical enumeration problem.

96: In chemistry and physics, SAWs and SAPs are used to model a variety of phenomena, including the 

97: properties of long-chain polymers in dilute solution \cite{deGennes},

98: the behaviour of ring polymers and vesicles in general \cite{LSF87} and 

99: benzenoid systems \cite{GCbook,VGJ02} in particular.

100: Though the qualitative form of the phase diagram \cite{FGW91} is known rigorously,

101: there is otherwise a paucity of rigorous results. However, there are a few conjectures, 

102: including the exact values of the critical exponents \cite{N82,N84}, 

103: and more recently the limit distribution of area and scaling function for SAPs, when 

104: enumerated by both area and perimeter \cite{RGJ01,C01,R02, RJG03, RJG04}.

105: 

106: Models of planar polygons with punctures arise naturally as cross-sections of

107: three-dimensional vesicle models. In such cross-sections, there may be holes

108: within holes, and the number of punctures may be infinite. In this work, we exclude

109: these possibilities. Whereas our methods can be used to study the former

110: case, the second situation presents new difficulties, which we have not yet 

111: overcome\footnote{Since punctured polygons with an unlimited number of punctures have, 

112: in contrast to polygons without punctures, an (ordinary) perimeter 

113: generating function with zero radius of convergence \cite{GJO01}, both the phase diagram 

114: and the detailed asymptotics are clearly going to

115: be very different from those of polygons without punctures. This is discussed 

116: further in the conclusion.}.

117: 

118: In this work we consider the effect of a finite number of punctures in polygon

119: models, in particular we study staircase polygons and self-avoiding polygons on the

120: square lattice. The perimeter of a punctured polygon \cite{GJWE00, JvRW89} is the 

121: perimeter of its boundary (both internal and external) while the area of a 

122: punctured polygon is the area of the enclosed by the external perimeter minus the 

123: area(s) of any holes\footnote{This has to 

124: be distinguished from so-called composite polygons \cite{J99}.

125: The perimeter of a composite polygon is defined as the perimeter of 

126: the external polygon only, resulting in asymptotic behaviour different

127: from punctured polygons. Moreover, composite polygons can have more complex 

128: internal structure than just other polygons.}.

129: As discussed in section \ref{sec:punc} below, the effect of punctures 

130: on the critical point and critical exponents of the area 

131: and perimeter generating function has been the subject of previous studies, but 

132: the effect of punctures on the critical amplitudes and detailed asymptotics

133: have not, to our knowledge, been previously considered.

134: 

135: Apart from the intrinsic interest of the problem, we also believe it to be the appropriate 

136: route to study the detailed asymptotics of {\it polyominoes}, 

137: since punctured polygons are a subclass of polyominoes.

138: While we still have some way to go to understand the polyomino phase diagram,

139: we feel that restricting the problem to this important subclass is the correct route. 

140: 

141: The make-up of the paper is as follows: In the next section we review the known situation 

142: for the perimeter and area generating functions of punctured polygons and polyominoes. 

143: In section 3 we review the phase diagram and scaling behaviour of staircase 

144: polygons and self-avoiding polygons. 

145: In section 4 we rigorously express the asymptotic behaviour of 

146: models of punctured polygons in the limit of large perimeter in terms of the 

147: asymptotic behaviour of the model without punctures, by refining arguments

148: used in \cite{GJWE00}. This leads, in particular, 

149: to a characterisation of the limit distribution of the area of punctured polygons. 

150: This result is then used to conjecture scaling functions of punctured polygons. 

151: We consider three cases of increasing generality. First, we 

152:  consider the case of minimal punctures. It is

153: shown that effects of self-avoidance are asymptotically irrelevant, and that elementary area 

154: counting arguments yield the leading asymptotic behaviour. We then discuss the 

155: case of a finite number of punctures of bounded size, and finally the case of a 

156: finite number of punctures of unbounded size. Results for the latter case

157: are given for models with a finite critical perimeter generating function

158: such as staircase polygons and self-avoiding polygons. Whereas the latter two 

159: cases are technically more involved, the underlying arguments are similar to the case

160: of minimal punctures. If the critical perimeter generating function of the

161: polygon model without punctures is finite, then all three cases lead, up to 

162: normalisation, to the same limit distributions 

163: and scaling function conjectures.

164: 

165: The next two sections discuss the development and application of 

166: extensive numerical data to test the results of the previous section. Moreover,

167: the numerical analysis yields predictions, conjectured to be exact, for the corrections to the

168: asymptotic behaviour. In particular, section 5 describes the very efficient algorithms 

169: used to generate the data, while section 6 applies a range of numerical tools to

170: the analysis of the generating functions for

171: punctured staircase polygons and then punctured self-avoiding polygons.

172: Here we wish to emphasise that our work on this problem

173: involved a close interplay between analytical and numerical work.

174: Initially, our intention was to check our predictions for scaling functions by

175: studying amplitude ratios for area moments (given in Table~\ref{tab:ratios}). 

176: We subsequently discovered numerically the {\em exact} solutions for minimally

177: punctured staircase polygons. We also obtained very accurate estimates for the 

178: amplitudes of staircase polygons with one or two punctures of arbitrary size.

179: From these results we were able to conjecture exact expressions for the amplitudes, 

180: which in turn spurred us on to further analytical work in order to prove these results.

181: The final section summarises and discusses our results.

182: 

183: \section{Punctured polygons \label{sec:punc}}

184: 

185: We consider polygons on the square lattice in this article. In particular, 

186: we study self-avoiding polygons and staircase polygons. 

187: A self-avoiding polygon on a lattice can be defined as a walk along

188: the edges of the lattice, which starts and ends at the same lattice point, but 

189: has no other self-intersections. When counting SAPs, they are generally considered 

190: distinct up to translations, 

191: change of starting point, and orientation of the walk, so if there are $p_m$ 

192: SAPs of length or perimeter $m$ there are $2mp_m$ walks (the factor of two 

193: arising since the walk can go in two directions). On the square lattice the perimeter

194: of any polygon is always even so it is natural to count polygons by half-perimeter

195: instead of perimeter. 

196: The area of a polygon is the number of lattice cells (times

197: the area of the unit cell) enclosed by the perimeter of the polygon.

198: A (square lattice) staircase polygon can be defined as the intersection of two mutually avoiding 

199: directed walks starting at the same lattice point, moving only to the right or up and 

200: terminating once the walks join at a vertex. Every staircase polygon is a self-avoiding 

201: polygon. It is well known that the number $p_m$ of staircase polygons of half-perimeter 

202: $m$ is given by the $(m-1)^{th}$ Catalan number, $p_m={2m-2 \choose m-1}/m$, with

203: half-perimeter generating function

204: %

205: \begin{equation}\label{eq:StGf}

206: \hspace{-2cm}

207: \PGf(x) =\sum_{m}p_m x^m= \frac{1-2x-\sqrt{1-4x}}{2} \sim \frac{1}{4}-

208: \frac{1}{2}(1-\mu x)^{2-\alpha} 

209: \qquad (\mu x\nearrow 1),

210: \end{equation}

211: %

212: where the connective constant $\mu=4$ and the critical exponent $\alpha=3/2$.

213: Recall that $f(x)\sim g(x)$ as $x\nearrow x_c$ means that $\lim f(x)/g(x)=1$ 

214: as $x\to x_c$ from below. In addition, as usual, the rhs is understood as the first

215: two leading terms in an asymptotic expansion of the lhs about $x=1/\mu$, see 

216: e.g.~\cite[Sec~1]{BH75}.

217: 

218: Punctured polygons \cite{GJWE00} are polygons with internal holes which are 

219: also polygons (the polygons are mutually- as well as self-avoiding).  The perimeter 

220: of a punctured polygon is the sum of the external and internal perimeters

221: while the area is the area of the external polygon {\em minus} the areas of the

222: internal polygons. We also consider polygons with {\em minimal} punctures, 

223: that is, polygons where the punctures are unit cells (or polygons with perimeter 4 

224: and area 1). Punctured staircase polygons are illustrated in figure~\ref{fig:poly}. 

225: %

226: \begin{figure}

227: \begin{center}

228: \includegraphics[scale=0.7]{punctured_stair}

229: \end{center}

230: \caption{\label{fig:poly} 

231: Examples of the types of staircase polygons we consider in this paper.

232: }

233: \end{figure}

234: 

235: We briefly review the situation for SAPs with punctures. Analogous results 

236: can be shown to hold for staircase 

237: polygons with punctures. Square lattice SAPs with $r$ punctures, counted by area $n$, 

238: were first studied by Janse van Rensburg and Whittington

239: \cite{JvRW89}. They proved the existence of an exponential growth constant 

240: $\kappa^{(r)}$ satisfying $\kappa^{(r)}=\kappa^{(0)}=\kappa$.

241: Denoting the corresponding number of SAPs by $a_n^{(r)}$ and assuming 

242: asymptotic behaviour of the form

243: %

244: \begin{displaymath}

245: a_n^{(r)} \sim A^{(r)}(\kappa^{(r)})^n n^{\beta_r-1}\qquad (n\to\infty),

246: \end{displaymath}

247: %

248: Janse van Rensburg proved \cite{JvR92} that $\beta_r = \beta_0 + r$. 

249: These results of course translate to the singular behaviour of the corresponding 

250: generating functions, defined by $\AGf^{(r)}(q) = \sum_{n > 0} a_n^{(r)}q^n$.

251: 

252: In \cite{GJWE00} Guttmann, Jensen, Wong and Enting studied square lattice 

253: SAPs with $r$ punctures counted by half-perimeter $m$. They proved the 

254: existence of an exponential growth constant $\mu^{(r)}$ satisfying

255: $\mu^{(r)}=\mu^{(0)}=\mu$. If the corresponding number $p_m^{(r)}$

256: of SAPs is assumed to behave asymptotically as 

257: %

258: \begin{displaymath}

259: p_m^{(r)}\sim B^{(r)}(\mu^{(r)})^m m^{\alpha_r-3}\qquad (m\to\infty),

260: \end{displaymath}

261: %

262: they argued, on the basis of a non-rigorous argument, that 

263: $\alpha_r = \alpha_0 + \frac32 r$. Their results also translate to the associated

264: half-perimeter generating function $\PP{r}(x) = \sum_{m > 0} p_m^{(r)}x^m$

265: correspondingly.

266: 

267: Similar results were obtained for polyominoes enumerated by number of cells 

268: (i.e. area) with a finite number $r$ of punctures \cite{GJWE00}. It has been 

269: proved that an exponential growth constant $\tau$ exists independently of $r$, 

270: which satisfies $4.06258\approx \tau >\kappa \approx 3.97087$, where $\kappa$ 

271: is the growth constant for SAPs enumerated by area. If the number

272: $a_n^{(r)}$ of polynominoes of area $n$ with $r$ punctures is assumed to 

273: satisfy asymptotically

274: %

275: \begin{displaymath}

276: a_n^{(r)} \sim C^{(r)}(\tau^{(r)})^n n^{\gamma_r-1}\qquad (n\to\infty),

277: \end{displaymath}

278: %

279: it has been shown that $\gamma_r=\gamma_0+r$ and hence that, if the exponents 

280: $\gamma_r$ exist, they increase by 1 per puncture. It was further conjectured

281: on the basis of extensive numerical studies \cite{GJWE00}, that the number

282: $a_n^{(r)}$ satisfies asymptotically

283: %

284: \begin{displaymath}

285: a_n^{(r)} \sim  \tau^n n^{r-1}\sum_{i \ge 0} C_i^{(r)}/n^{i}

286: \qquad (n\to\infty).

287: \end{displaymath}

288: %

289: Notice the conjecture $\gamma_0=0$ and that the correction terms 

290: go down by a whole power.

291: 

292: For unrestricted polyominoes, that is to say, with no restriction on the number 

293: of punctures, it was proved by Guttmann, Jensen and Owczarek \cite{GJO01} that 

294: the  perimeter generating function has zero radius of convergence.

295: The perimeter is defined to be the perimeter of the boundary plus the

296: total perimeter of any holes.

297: If $p_{m}$  denotes the number of polyominoes, distinct up to a translation,

298: with  half-perimeter $m$, they proved that $p_{m} = m^{m/4 + {\rm o}(m)}$,

299: meaning that

300: % 

301: $$\lim_{m \to \infty}\frac {\log p_{m}}{m \log {m}} = \frac{1}{4}.$$

302: %

303: An attempt to study the quasi-exponential generating function with coefficients

304: $r_m = p_m/\Gamma(m/4+1)$ was equivocal.

305: For that reason, studying punctured  self-avoiding polygons was 

306: considered a controlled route to attempt to

307: determine the two-variable area-perimeter generating function of polyominoes.

308: 

309: In passing, we note that in \cite{GJ06b} the exact solution of the perimeter generating 

310: function for staircase polygons with a staircase hole is conjectured, in the form of 

311: an $8^{th}$ order ODE. It is not obvious how to extract particular asymptotic information, 

312: notably critical amplitudes from the solution without numerically integrating the ODE. 

313: In the following, we will obtain such information by combinatorial arguments, 

314: which refine those of \cite{GJWE00}.

315: 

316: 

317: \section{Polygon models and their scaling behaviour}

318: 

319: We review the asymptotic behaviour of self-avoiding polygons

320: and staircase polygons following mainly \cite{FGW91}. 

321: For concreteness, consider the fixed perimeter ensemble where, for 

322: fixed half-perimeter $m$, each polygon of area $n$ has 

323: a weight proportional to $q^n$, for some positive real number $q$. 

324: If $0<q<1$, polygons of large area are exponentially suppressed, 

325: so that typical polygons should be ramified objects. Since such 

326: polygons would closely resemble branched polymers, the phase $0<q<1$ 

327: is also referred to as the {\it branched polymer phase}. As $q$ 

328: approaches unity, typical polygons should fill out more, and 

329: become less string-like. For $q>1$, polygons of small area are 

330: exponentially suppressed, so that typical polygons should become 

331: ``fat''. Indeed, they resemble convex polygons \cite{OP99a} and

332: it has been proved \cite{FGW91} that the mean area of polygons of 

333: half-perimeter $m$ grows asymptotically proportional to $m^2$.

334: In the extended phase $q=1$, it is numerically very well established that 

335: the mean area of polygons of half-perimeter $m$ grows 

336: asymptotically proportionally to $m^{3/2}$. In the branched polymer 

337: phase $0<q<1$, the mean area of polygons of half-perimeter $m$ is 

338: expected to grow asymptotically linearly in $m$, compare also 

339: \cite[Thm 7.6]{J00} and \cite[Ch~IX.6, Ex.~12]{FS06}.

340: 

341: This change of asymptotic behaviour of typical polygons w.r.t.~$q$ 

342: is reflected in the singular behaviour of the half-perimeter and area 

343: generating function

344: %

345: \begin{displaymath}

346: \PGf(x,q)=\sum_{m,n} p_{m,n} x^m q^n,

347: \end{displaymath}

348: %

349: where $p_{m,n}$ denotes the number of (self-avoiding) polygons of 

350: half-perimeter $m$ and area $n$. It has been proved \cite{FGW91} 

351: that the free energy

352: %

353: \begin{displaymath}

354:  \kappa (q):= \lim_{m \to \infty} \frac{1}{m} \log \left(\sum_n p_{m,n} q^n\right) 

355: \end{displaymath}

356: %

357: exists and is finite if $0<q \le 1$. Further, $\kappa(q)$ is log-convex and continuous for these 

358: values of $q$. It is infinite for $q>1$.

359: It was proved that for fixed $0<q\le1$, the radius of convergence $x_c(q)$ of $\PGf(x,q)$ 

360: is given by $x_c(q)= e^{-\kappa(q)}$.

361: For fixed $q>1$, $\PGf(x,q)$ has zero radius of convergence.

362: Fisher et al. \cite{FGW91} obtained rigorous upper and

363: lower bounds on $x_c(q)$.

364: The expected phase diagram, i.e., the radius of convergence of $\PGf(x,q)$

365: in the $x-q$ plane, as estimated numerically from extrapolation of SAP 

366: enumeration data by perimeter and area, 

367: is sketched qualitatively in figure~\ref{fig:phase}.

368: %

369: \begin{figure}[htbp]

370:   \centering

371:   \includegraphics[width=5cm]{phase}

372:   \caption{\label{fig:phase} A sketch of the phase diagram of self-avoiding polygons.}

373: \end{figure}

374: 

375: For $0<q<1$, the line $x_c(q)$ is, for self-avoiding polygons, expected to be a 

376: line of logarithmic singularities of the generating function $\PGf(x,q)$. 

377: For branched polymers in the continuum limit, the existence of the logarithmic 

378: singularity has recently been proved \cite{BI03}.

379: The line $q=1$ is, for $0<x<x_c:=x_c(1)$, a line of finite essential singularities 

380: \cite{FGW91}. For staircase polygons, counted by half-perimeter and area, the 

381: corresponding phase diagram can be determined exactly, and is qualitatively 

382: similar to that of self-avoiding polygons. Along the line $x_c(q)$ the 

383: half-perimeter and area generating function diverges with a simple pole, and 

384: the line $q=1$ is, for $0<x<x_c$, a line of finite essential 

385: singularities \cite{Pr94}. 

386: 

387: \medskip 

388: 

389: We will focus on the uniform fixed perimeter ensemble $q=1$

390: in this article.

391: Whereas asymptotic area laws in the fixed perimeter ensemble are

392: expected to be Gaussian for positive $q\ne1$, the behaviour in the uniform 

393: fixed perimeter ensemble $q=1$ is more interesting. For 

394: staircase polygons, it can be shown that a limit distribution of 

395: area exists and is given by the Airy distribution \cite{L84,T91,FL01}. 

396: For self-avoiding polygons, it is conjectured that an area limit law exists and

397: is given by the Airy distribution, on the basis of a detailed numerical analysis 

398: \cite{RGJ01,RJG03,RJG04}. See subsections  \ref{sec:scaling_min} and

399: \ref{sec:scaling_func}.

400: 

401: If $p_{m,n}$ denotes the number of polygons of half-perimeter $m$ and area $n$,

402: the existence and the form of a limit distribution can be inferred from the asymptotic

403: behaviour of the factorial moment coefficients $\sum_{n}(n)_k p_{m,n}$,

404: where $(a)_k=a\cdot(a-1)\cdot\ldots\cdot(a-k+1)$. The following

405: result is obtained by standard reasoning \cite{C74}.

406: 

407: \begin{prop}\label{theo:general}

408: Let for $m,n\in\mathbb N_0$ real numbers $p_{m,n}$ be given.

409: Assume that the numbers $p_{m,n}$ have the asymptotic form, for $k\in\mathbb N_0$,

410: %

411: \begin{equation}\label{Ecr:ampli}

412: \sum_n (n)_k p_{m,n}\sim A_k x_c^{-m} 

413: m^{\gamma_k-1}\qquad (m\to\infty)

414: \end{equation}

415: %

416: for positive real numbers $A_k$ and $x_c$, where $\gamma_k=(k-\theta)/\phi$, 

417: with real constants $\theta$ and 

418: $\phi>0$. Assume that the numbers $M_k:=A_k/A_0$ satisfy

419: the Carleman condition

420: %

421: \begin{equation}\label{Ecr:growth}

422: \sum_{k=0}^\infty (M_{2k})^{-1/{2k}}=+\infty.

423: \end{equation}

424: %

425: Then, for almost all $m$, the random variables $\widetilde X_m$

426: of area in the uniform fixed perimeter ensemble

427: %

428: \begin{displaymath}

429: \mathbb P(\widetilde X_m=n)=\frac{p_{m,n}}{\sum_np_{m,n}}

430: \end{displaymath}

431: %

432: are well defined. We have

433: %

434: \begin{displaymath}

435: X_m:=\frac{\widetilde X_m}{m^{1/\phi}} \stackrel{d}{\longrightarrow}X \qquad (m\to\infty),

436: \end{displaymath}

437: %

438: for a uniquely defined random variable $X$ with moments $M_k$,

439: where the superscript $\stackrel{d}{}$ denotes convergence in distribution.

440: We also have moment convergence.

441: \end{prop}

442: 

443: \begin{proof}[Sketch of proof]

444: A straightforward calculation using Eq.~(\ref{Ecr:ampli}) leads to

445: %

446: \begin{displaymath}

447: \mathbb E[(\widetilde X_m)_k]

448: \sim \frac{A_k}{A_0} m^{k/\phi}

449: \qquad (m\to\infty).

450: \end{displaymath}

451: %

452: It follows that asymptotically 

453: the factorial moments are equal to the (ordinary) moments.

454: Thus, the moments of $X_m$ have the same asymptotic form

455: %

456: \begin{displaymath}

457: \mathbb E[(X_m)^k]

458: \sim\frac{A_k}{A_0}=M_k\qquad (m\to\infty).

459: \end{displaymath}

460: %

461: Due to the growth condition Eq.~(\ref{Ecr:growth}), the sequence 

462: $(M_k)_{k\in\mathbb N_0}$ defines a unique random variable $X$ with moments $M_k$.

463: Moment convergence of $(X_m)$ implies convergence in distribution,

464: see \cite[Thm 4.5.5]{C74} for the line of arguments.

465: \end{proof}

466: 

467: \medskip

468: 

469: The assumption Eq.~(\ref{Ecr:ampli}) 

470: translates, on the level of the half-perimeter and area generating function

471: $\PGf(x,q)$, to a certain asymptotic

472: behaviour of the so-called factorial moment generating functions

473: %

474: \begin{displaymath}

475: g_k(x)=\frac{(-1)^k}{k!}\left.\frac{\partial^k}{\partial q^k}

476: \PGf(x,q)\right|_{q=1}.

477: \end{displaymath}

478: %

479: It can be shown (compare \cite{F99}) that the asymptotic behaviour Eq.~(\ref{Ecr:ampli}) 

480: implies for $\gamma_k>0$ the asymptotic equivalence

481: %

482: %

483: \begin{equation}\label{eq:relation2}

484: g_k(x)\sim \frac{f_k}{(x_c-x)^{\gamma_k}}\qquad (x\nearrow x_c),

485: \end{equation}

486: %

487: where the amplitudes $f_k$ are related to the amplitudes $A_k$%

488: \footnote{

489: Note that our definition of the amplitudes $A_k$ differs

490: from that in \cite{R02} by a factor of $(-1)^kk!$ and from

491: that in \cite{R05,R06} by a factor of $k!$.} 

492: in Proposition~\ref{theo:general} via 

493: %

494: \begin{equation}\label{eq:relation}

495: f_k=\frac{(-1)^k}{k!}A_kx_c^{\gamma_k}\Gamma(\gamma_k).

496: \end{equation}

497: %

498: If $-1<\gamma_k<0$, the series $g_k(x)$ is convergent as $x\nearrow x_c$, and

499: the same estimate Eq.~(\ref{eq:relation2}) holds, 

500: with $g_k(x)$ replaced by $g_k(x)-g_k(x_c)$, where 

501: $g(x_c):=\lim_{x\nearrow x_c} g(x)$.

502: In order to deal with these two different cases, we define for 

503: a power series $g(x)$ with radius of convergence $x_c$, the number

504: %

505: \begin{displaymath}

506: g^{(c)}=\left\{

507: \begin{array}{cc}

508: g(x_c)& {\rm if} \,\,|\lim_{x\nearrow x_c} g(x)|<\infty\\

509: 0 & {\rm otherwise}.

510: \end{array}

511: \right.

512: \end{displaymath}

513: %

514: Adopting the generating function

515: point of view, the amplitudes $f_k$ determine the numbers $A_k$ and

516: hence the moments $M_k=A_k/A_0$ of the limit distribution.

517: The formal series $F(s)=\sum_{k\ge0} f_k s^{-\gamma_k}$ will appear

518: frequently in the sequel.

519: 

520: \begin{definition}\label{def:fscl}

521: %

522: For the generating function $\PGf(x,q)$ of a class of

523: self-avoiding polygons, denote its factorial moment 

524: generating functions by

525: %

526: \begin{displaymath}

527: g_k(x)=\frac{(-1)^k}{k!}\left.\frac{\partial^k}{\partial q^k}

528: \PGf(x,q)\right|_{q=1}.

529: \end{displaymath}

530: %

531: Assume that the factorial moment generating functions satisfy

532: %

533: \begin{equation}\label{form:approx}

534: g_k(x)-g_k^{(c)} \sim \frac{f_k}{(x_c-x)^{\gamma_k}} \qquad (x\nearrow x_c),

535: \end{equation}

536: %

537: with real exponents $\gamma_k$. Then, the formal series

538: %

539: \begin{displaymath}

540: F(s) = \sum_{k\ge0}f_k s^{-\gamma_k}

541: \end{displaymath}

542: %

543: is called the area amplitude series.

544: \end{definition}

545: 

546: The area amplitude series is expected to approximate

547: the half-perimeter and area generating function $\PGf(x,q)$ about 

548: $(x,q)=(x_c,1)$. This is motivated by the following heuristic 

549: argument. Assume that $\gamma_k=(k-\theta)/\phi$ with $\phi>0$ and argue

550: %

551: \begin{eqnarray*}

552: \PGf(x,q)&\approx \sum_{k\ge0} 

553: \left(g_k^{(c)}+\frac{f_k}{(x_c-x)^{\gamma_k}}\right)(1-q)^k\\

554: &\approx \left(\sum_{k\ge0}g_k^{(c)}(1-q)^k\right)+(1-q)^{\theta}\left(\sum_{k\ge0}

555: f_k \left( \frac{x_c-x}{(1-q)^{\phi}}\right)^{-\gamma_k}

556: \right).

557: \end{eqnarray*}

558: %

559: In the above calculation, we formally expanded $\PGf(x,q)$

560: about $q=1$ and then replaced the Taylor coefficients by their

561: leading singular behaviour about $x=x_c$. 

562: In the rhs of the above expression, the first sum is by assumption finite,

563: and the second term contains the area amplitude

564: series $F(s)$ of combined argument $s=(x_c-x)/(1-q)^\phi$.

565: This motivates the following definition. A class of self-avoiding

566: polygons is a subset of self-avoiding polygons. Prominent examples are, 

567: among others \cite{BM96}, self-avoiding polygons and staircase polygons.

568: 

569: \begin{definition}\label{def:sf}

570: Let a class of square lattice self-avoiding polygons be given, with

571: half-perimeter and area generating function $\PGf(x,q)$. 

572: Let $0<x_c<\infty$ be the radius of convergence 

573: of the half-perimeter generating function $\PGf(x,1)$. Assume that there 

574: exist a constant $s_0\in[-\infty,0)$, a function 

575: ${\cal F}:(s_0,\infty)\to\mathbb R$, a real constant $A$ and real

576: numbers $\theta$ and $\phi>0$, such 

577: that the generating function $\PGf(x,q)$ satisfies, for real 

578: $x$ and $q$, where $0<q<1$ and $(x_c-x)/(1-q)^\phi\in(s_0,\infty)$,

579: the asymptotic equivalence 

580: %

581: \begin{equation}\label{form:scaling}

582: \PGf(x,q)\sim A+(1-q)^\theta {\cal F}\left(\frac{x_c-x}{(1-q)^{\phi}}\right)

583: \qquad (x,q)\longrightarrow (x_c,1).

584: \end{equation}

585: %

586: Then, the function ${\cal F}(s)$  is called a {\it scaling 

587: function} of combined argument 

588: $s=(x_c-x)/(1-q)^{\phi}$, and $\theta$ and $\phi$ 

589: are called {\it critical exponents}.

590: \end{definition}

591: 

592: \noindent {\bf  Remarks.} {\it i)}

593: Due to the restriction on the argument of the scaling function, the 

594: limit $(x,q)\to(x_c,1)$ is approached for values $(x,q)$ satisfying

595: $x<x_0(q)$ and $q<1$, where $x_0(q)= x_c-s_0(1-q)^\phi$. 

596: %If $s_0$ is a singularity of $F(s)$, the curve $x_0(q)$ describes the 

597: %radius of convergence of $\PGf(x,q)$ for $q<1$ near $q=1$. 

598: \\

599: {\it ii)} The above scaling form is also suggested by the theory 

600: of tricritical scaling, adapted to polygon models \cite{BOP93}. The 

601: scaling function describes the leading singular behaviour

602: of $\PGf(x,q)$ about the point $(x_c,1)$ where the two lines of 

603: qualitatively different singularities meet. \\

604: {\it iii)} The additional condition $\phi>\theta$ and 

605: $\theta\notin \mathbb N_0$ ensure that $\gamma_k\in(-1,\infty)

606: \setminus\{0\}$.

607: Then, by the above argument, it is plausible that there exists

608: an asymptotic expansion of the scaling function ${\cal F}(s)$ 

609: about infinity coinciding with the area amplitude series 

610: $F(s)$, i.e., ${\cal F}(s)\sim F(s)$ as $s\to\infty$. Recall that

611: $s$ is considered to be a real parameter.

612: 

613: \medskip

614: 

615: For staircase polygons the existence of a scaling form 

616: Eq.~(\ref{form:scaling}) has been proved \cite[Thm~5.3]{Pr94},

617: with scaling function ${\cal F}(s):(s_0,\infty)\to\mathbb R$ 

618: explicitly given by

619: %

620: \begin{equation}\label{sfstair}

621: {\cal F}(s) = \frac{1}{16}\frac{\rm d}{{\rm d}s}\log\mbox{Ai} 

622: \left(  2^{8/3} s \right),

623: \end{equation}

624: %

625: with exponents $\theta=1/3$ and $\phi=2/3$ and $x_c=1/4$, where

626: $\mbox{Ai}(x)=\frac{1}{\pi}\int_0^\infty \cos(t^3/3+tx){\rm d}t$ is the

627: Airy function. The constant $s_0$ is such that $2^{8/3}s_0$ is the

628: location of the Airy function zero of smallest modulus.  For {\it rooted} 

629: SAPs with half-perimeter and area generating 

630: function $\PGf_r(x,q)=x\frac{{\rm d}}{{\rm d}x}\PGf(x,q)$, the conjectured form of 

631: the scaling function ${\cal F}_r(s):(s_0,\infty)\to\mathbb R$ is \cite{R02}

632: %

633: \begin{displaymath}

634: {\cal F}_r(s) = \frac{x_c}{2\pi}\frac{\rm d}{{\rm d}s}\log\mbox{Ai} \left( \frac{\pi}{x_c} 

635: \left( 4A_0\right)^\frac{2}{3} s \right),

636: \end{displaymath}

637: %

638: with the same exponents as for staircase polygons, $\theta=1/3$ 

639: and $\phi=2/3$. Here, $x_c=0.14368062927(2)$ is the radius of convergence

640: of the half-perimeter generating function $\PGf_r(x,1)$ of (rooted) SAPs, and 

641: $A_0=0.09940174(4)$ is the critical amplitude $\sum_n m p_{m,n}\sim A_0 

642: x_c^{-m}m^{-3/2}$ of rooted SAPs, which coincides with the critical amplitude

643: $A_0$ of (unrooted) SAPs. Again, the constant $s_0$ is such that the corresponding 

644: Airy function argument is the location of the Airy function zero of smallest modulus.

645: This conjecture was based on the conjecture that both models

646: have, up to normalisation constants, the same area amplitude series. 

647: The latter conjecture is supported numerically  

648: to very high accuracy by an extrapolation of the moment series using exact 

649: enumeration data \cite{RGJ01,RJG03}. The conjectured form of the scaling 

650: function ${\cal F}(s):(s_0,\infty)\to\mathbb R$ for SAPs is obtained 

651: by integration,

652: %

653: \begin{equation}\label{form:unroot}

654: {\cal F}(s) = -\frac{1}{2\pi} \log\mbox{Ai} 

655: \left( \frac{\pi}{x_c} \left( 4A_0\right)^\frac{2}{3} s \right)+C(q),

656: \end{equation}

657: %

658: with exponents $\theta=1$ and $\phi=2/3$. In the above formula, $C(q)$ 

659: is a $q$ dependent constant of integration, $C(q) = \frac{1}{12 \pi}

660: (1 - q)\log(1 - q)$, see \cite{RJG04}.

661: Corresponding results for the triangular and hexagonal lattices can be found in \cite{RGJ01}.

662: 

663: For models of punctured polygons with a finite number of punctures, we have 

664: qualitatively the same phase diagram as for polygon models without punctures, 

665: however with different critical exponents $\theta$ depending on the number of 

666: punctures \cite{JvR92,GJWE00}, and hence we expect different scaling functions. 

667: We will focus on critical exponents and area limit laws in the uniform

668: ensemble $q=1$ in the following section. This will lead to conjectures for the

669: corresponding scaling functions.

670: 

671: \section{Scaling behaviour of punctured polygons \label{sec:scaling}}

672: We briefly preview the main results of this section.

673: In subsection~\ref{sec:scaling_min} we study polygons with a finite number

674: of minimal punctures.  Our result assumes a certain asymptotic form for 

675: the area moment coefficients for unpunctured polygons. This `assumed' form is 

676: known to be true for staircase polygons and many other models and universally 

677: accepted as true for self-avoiding polygons. Given this assumption, we prove 

678: that the asymptotic behaviour of the area moment coefficients for minimally 

679: punctured polygons can be expressed in terms of the asymptotic behaviour of 

680: unpunctured polygons.

681: In particular we derive expressions for the leading amplitude of the area moments 

682: for punctured polygons in terms of the amplitudes for unpunctured

683: polygons. For staircase polygons this leads to exact formulas for the amplitudes.

684: For self-avoiding polygons the formulas contain certain constants which aren't

685: known exactly but can be estimated numerically to a very high degree of accuracy.

686: In subsection~\ref{sec:scaling_bound} we extend the study and proofs to 

687: polygons with a finite number of punctures of {\em bounded} size and then

688: in subsection~\ref{sec:scaling_arbitrary} to models with punctures of arbitrary

689: or unbounded size. Finally in subsection~\ref{sec:scaling_func} we consider

690: the consequences of our results for the area limit laws of punctured polygons

691: and we present conjectures for the scaling functions.

692: 

693: \subsection{Polygons with $r$ minimal punctures \label{sec:scaling_min}}

694: 

695: For polygon models with rational perimeter generating functions,

696: corresponding models with minimal punctures have been studied in 

697: \cite{RG01}. In particular, 

698: a method to derive explicit expressions for generating functions

699: of exactly solvable models with a minimal puncture was 

700: given \cite[Appendix]{RG01}. It has been applied to Ferrers

701: diagrams, whose perimeter and area generating function satisfies 

702: a linear $q$-difference equation, see \cite[Eq.~(54)]{RG01}. 

703: The method can also be applied to the model of staircase polygons, 

704: whose half-perimeter and area generating function ${\cal P}(x,q)$

705: satisfies the quadratic $q$-difference equation

706: %

707: \begin{equation}\label{sp0}

708: {\cal P}(x,q)=\frac{x^2q}{1-2qx-{\cal P}(qx,q)}.

709: \end{equation}

710: %

711: Let $\StM{r}(x,q)$ denote the half-perimeter and area generating function

712: of staircase polygons with $r$ minimal punctures. We have the following 

713: result for the case $r=1$.

714: %

715: \begin{fact}\label{fact}

716: The half-perimeter and area generating function of staircase polygons with a 

717: single minimal puncture $\StM{1}(x,q)$ is given by

718: %

719: \begin{equation}\label{sp1}

720: \StM{1}(x,q)=\frac{x^4}{(1-2qx-{\cal P}(qx,q))^2} 

721: \left( {\cal P}(qx,q)-qx \frac{\partial{\cal P} }{\partial x}(qx,q)+

722: q \frac{\partial {\cal P}}{\partial q}(qx,q) \right),

723: \end{equation}

724: %

725: where ${\cal P}(x,q)$ satisfies Eq.~(\ref{sp0}).

726: \qed

727: \end{fact}

728: %

729: \noindent {\bf Remarks.} 

730: %

731: {\it i)} For a proof of Fact \ref{fact}, proceed along the lines of 

732: \cite[Appendix]{RG01}. We do not give the details, since we are mainly 

733: interested in asymptotic results, for which we will give an elementary

734: combinatorial derivation, valid for arbitrary $r$. See 

735: Proposition \ref{prop:minpunc} and its subsequent extensions.\\

736: %

737: {\it ii)} For polygons with $r$ punctures, their {\bf $k^{th}$} area moment generating 

738: functions are defined by $\StM{r}_k(x)= \left(q\frac{\partial}

739: {\partial q}\right)^k \left.\StM{r}(x,q) \right|_{q=1}$. The above equations can be used 

740: to obtain explicit expressions for the area moment generating functions $\StM{1}_k(x)$ 

741: by implicit differentiation. The functions $\StM{1}_k(x)$ also appear in

742: section \ref{sec:stair_min}.\\

743: %

744: {\it iii)} Assuming that $\StM{1}(x,q)$ has scaling behaviour of the form

745: %

746: \begin{displaymath}

747: \StM{1}(x,q) \sim (1-q)^{\theta_1}\FM{1}((x_c-x)(1-q)^{-\phi_1})

748: \end{displaymath}

749: %

750: about $(x,q)=(x_c,1)$, and the necessary analyticity conditions for the

751: validity of the following calculation, we can 

752: express the scaling function $\FM{1}(s)$ of staircase polygons with 

753: a single minimal puncture in terms of the

754: known scaling function ${\cal F}(s)$ of staircase polygons Eq.~(\ref{sfstair}). 

755: From Eq.~(\ref{sp1}) we infer that $\theta_1=-2/3$, $\phi_1=2/3$ and

756: %

757: \begin{equation}\label{stsf}

758: \FM{1}(s)=\frac{1}{24}s{\cal F}'(s)-\frac{1}{48} {\cal F}(s).

759: \end{equation}

760: 

761: \medskip

762: 

763: In principle, the method of \cite[Appendix]{RG01} can 

764: be used to analyse the case of several minimal punctures. 

765: However, the analysis becomes quite cumbersome. On the other 

766: hand, the previous result suggests simple expressions for 

767: the scaling functions of models with several punctures in terms 

768: of that without a puncture. Moreover, we expect such

769: a phenomenon also to occur for models where an exact 

770: solution does not exist or is not known. This is 

771: discussed next. We will asymptotically analyse the area 

772: moments of a polygon model with punctures and

773: draw conclusions about their possible scaling behaviour.

774: 

775: For a class of punctured self-avoiding 

776: polygons, consider their area moment coefficients

777: %

778: \begin{displaymath}

779: p_m^{\Box(r,k)}:= \sum_n n^kp_{m,n}^{\Box(r)},

780: \end{displaymath}

781: %

782: where $p_{m,n}^{\Box(r)}$ denotes the number of polygons 

783: in the class  with $r$ minimal punctures, $r\in\mathbb N_0$, 

784: of half-perimeter $m$ and area $n$. For simplicity of notation,

785: we write $p_{m,n}:=p_{m,n}^{\Box(0)}$ and $p_m^{(k)}:=

786: p_{m}^{\Box(0,k)}$. The area moments in the uniform fixed

787: perimeter ensemble are expressed in terms of 

788: the area moment coefficients via

789: %

790: \begin{equation}

791: \mathbb E[(\widetilde X_m^{\Box(r)})^k]=\frac{\sum_n n^k p_{m,n}^{\Box(r)}}

792: {\sum_n p_{m,n}^{\Box(r)}} = \frac{p_m^{\Box(r,k)}}{p_m^{\Box(r,0)}}.

793: \end{equation}

794: %

795: 

796: \medskip

797: 

798: \begin{prop}\label{prop:minpunc}

799: Assume that, for a class of self-avoiding polygons without punctures, 

800: the area moment coefficients $p_m^{(k)}$ have

801: the asymptotic form, for $k\in\mathbb N_0$,

802: %

803: \begin{equation}\label{pmas}

804: p_m^{(k)} \sim A_k x_c^{-m} m^{\gamma_k-1} \qquad (m\to\infty),

805: \end{equation} 

806: %

807: for numbers $A_k>0$, $x_c>0$ and exponents $\gamma_k=(k-\theta)/\phi$, 

808: where $\theta$ and $\phi$ are real constants and $0<\phi<1$. 

809: Then, the area moment coefficient $p_m^{\Box(r,k)}$ of the polygon class

810: with $r\ge1$ minimal punctures is asymptotically given by, for $k\in\mathbb N_0$, 

811: %

812: \begin{equation}\label{form:minprk}

813: p_m^{\Box(r,k)} \sim A_k^{(r)} x_c^{-m} m^{\gamma_k^{(r)}-1} \qquad (m\to\infty),

814: \end{equation}

815: where $A_k^{(r)}=A_{k+r}x_c^{2r}/r!$ and $\gamma_k^{(r)}=\gamma_{k+r}$. 

816: \end{prop}

817: 

818: \begin{proof}

819: We will derive upper and lower bounds on $p_m^{\Box(r,k)}$, which will be 

820: shown to coincide asymptotically. Let us call two polygons interacting if 

821: their boundary curves have non-empty intersection. An upper bound is obtained 

822: by allowing for interaction between all constituents of a punctured polygon. 

823: Let a polygon $\Pol$ of half-perimeter $m-2r$ and area $n+r$ be given. The 

824: number of ways of placing $r$ squares inside $\Pol$ is clearly less than

825: $(n+r)^r/r!$. We thus have

826: %

827: \begin{displaymath}

828: p_m^{\Box(r,k)}\le {\widetilde p}_m^{(r,k)} := \frac{1}{r!}

829: \sum_{n\ge 1} n^k (n+r)^rp_{m-2r,n+r} =\frac{1}{r!}

830: \sum_{n\ge r+1} (n-r)^k n^rp_{m-2r,n}.

831: \end{displaymath}

832: %

833: By Bernoulli's inequality, we get for ${\widetilde p}_m^{(r,k)}$ the 

834: bound

835: %

836: \begin{displaymath}

837: \frac{1}{r!} \sum_{n\ge r+1} \left( n^{k+r}-kr\,n^{k+r-1}\right) 

838: p_{m-2r,n}\le {\widetilde p}_m^{(r,k)}\le \frac{1}{r!} 

839: \sum_{n\ge r+1} n^{k+r}p_{m-2r,n}.

840: \end{displaymath}

841: %

842: For every polygon of perimeter $2s$ and area $t$ we 

843: have $t\ge s-1$. Thus, for $m$ sufficiently large, we can 

844: replace the lower bound of summation $r+1$ by zero.

845: In particular, the latter relation is for $m\ge 3r+2$  

846: equivalent to

847: %

848: \begin{displaymath}

849: \frac{1}{r!} \left( p_{m-2r}^{(k+r)}-kr\,p_{m-2r}^{

850: (k+r-1)} \right)\le {\widetilde p}_m^{(r,k)}\le 

851: \frac{1}{r!} p_{m-2r}^{(k+r)}.

852: \end{displaymath}

853: %

854: The assumption Eq.~(\ref{pmas}) on the asymptotic 

855: behaviour of $p_{m}^{(k)}$ then implies that

856: %

857: \begin{displaymath}

858: {\widetilde p}_m^{(r,k)}\sim \frac{x_c^{2r}}{r!}p_m^{

859: (k+r)}\qquad (m\to\infty).

860: \end{displaymath}

861: %

862: We derive a lower bound by subtracting from the upper bound an 

863: upper bound on the number of square-square and square-boundary 

864: interactions. Clearly, square-square interactions are only

865: present for $r>1$. For a given polygon $\Pol$, the number 

866: of square-square interactions of $r$ squares is smaller than the number of 

867: interactions between two squares, where the remaining 

868: $r-2$ squares may occur at arbitrary positions within the polygon. There

869: are five possible configurations for an interaction between two

870: squares, yielding the upper bound $5(n+r)(n+r)^{r-2}$. Thus, the

871: contribution to ${\widetilde p}_{m}^{(r,k)}$ from square-square

872: interactions is bounded from above by

873: %

874: \begin{displaymath}

875: \sum_{n\ge 1} n^k 5(n+r)(n+r)^{r-2}p_{m-2r,n+r}=

876: 5(r-1)!{\widetilde p}_{m}^{(r-1,k)},

877: \end{displaymath}

878: %

879: which is asymptotically negligible compared to ${\widetilde p}_{m}

880: ^{(r,k)}$. Similarly, the number of configurations arising from

881: square-boundary interactions is bounded from above by $\sum_{j=1}^r 

882: 4^j(m-2r)^j(n+r)^{r-j}$. This bound is obtained by estimating the 

883: number of configurations of $j$ squares at the boundary by 

884: $4^j(m-2r)^j$, the factor 4 arising from edge and vertex interactions,

885: the factor $(n+r)^{r-j}$ accounting for arbitrary positions of the

886: remaining $(r-j)$ squares. We thus get an upper bound

887: %

888: \begin{eqnarray*}

889: \sum_{j=1}^r 4^j (m-2r)^j (r-j)! {\widetilde p}_m^{(r-j,k)}

890: \sim \sum_{j=1}^r 4^j x_c^{2r-2j}(m-2r)^j p_m^{(k+r-j)}\\  

891:  \sim 4 x_c^{2r-2}m\, p_m^{(k+r-1)}

892: \qquad (m\to\infty).

893: \end{eqnarray*}

894: %

895: By assumption, the latter bound is asymptotically negligible compared to 

896: ${\widetilde p}_{m}^{(r,k)}$. Thus, the lower bound is asymptotically

897: equal to the upper bound, which yields the assertion of the proposition.

898:  \end{proof}

899: 

900: \noindent {\bf Remarks.} 

901: %

902: {\it i)} Proposition \ref{prop:minpunc} expresses the asymptotic

903: behaviour of the area moment coefficients of minimally 

904: punctured polygons in terms of those of polygons without punctures.

905: The assumption Eq.~(\ref{pmas}) on the growth of the area moment coefficients 

906: of the model without punctures is satisfied for the

907: usual polygon models \cite{BM96}. The asymptotic behaviour of some models 

908: satisfying $\phi=1$, to which Proposition \ref{prop:minpunc} does not 

909: apply, has been studied in \cite{RG01}.\\

910: \noindent {\it ii)}

911: As discussed in the previous subsection, the amplitudes 

912: $A_k$ are related to the amplitudes $f_k$ of Eq.~(\ref{form:approx}) by

913: Eq.~(\ref{eq:relation}), if $\gamma_k\in(-1,\infty)\setminus\{0\}$. 

914: For staircase polygons, where $\theta=1/3$ 

915: and $\phi=2/3$, we have explicit expressions for the

916: amplitudes $A_k$. More generally, it has been shown \cite{R02,R05,R06}

917: that, for classes of polygon models whose generating function satisfies a 

918: $q$-functional equation with a square root as the dominant singularity

919: of their perimeter generating function, we have $f_k=c_kf_1^kf_0^{1-k}$,

920: where the numbers $c_k$ are, for $k\ge1$, given by

921: %

922: \begin{equation}\label{ck}

923: \gamma_{k-1}c_{k-1}+\frac{1}{4}\sum_{l=0}^{k}c_{k-l}c_l=0, \qquad c_0=1.

924: \end{equation}

925: %

926: The critical point $x_c$ as well as $f_0$ and $f_1$ 

927: are model dependent constants. For staircase polygons we have 

928: $x_c=1/4$, $f_0=-1$ and $f_1=-1/64$.\\

929: \noindent {\it iii)} Rooted self-avoiding polygons are conjectured to also have

930: the exponents $\theta=1/3$ and $\phi=2/3$. In this case the asymptotic form 

931: Eq.~(\ref{pmas}) and the form of the amplitudes $A_k$, given in 

932: Eqs.~(\ref{eq:relation}) and (\ref{ck}), has been tested for $k\le10$ 

933: and shown to hold for  to a high degree of 

934: numerical accuracy \cite{RJG03}. Here $x_c=0.14368062927(2)$ is the

935: radius of convergence of the (rooted) SAP half-perimeter generating function,

936: $f_0=-0.929607(1)$ and $f_1=-x_c/(8\pi)$ are the rooted SAP critical amplitudes

937: as in Eq.~(\ref{form:approx}). We conjecture that the asymptotic form~(\ref{pmas}) 

938: holds for rooted SAPs for all values of $k$. Accepting this conjecture to be true, 

939: Proposition~\ref{prop:minpunc} gives the asymptotic behaviour

940: for rooted self-avoiding polygons with $r$ minimal punctures.

941: By definition, unrooted SAPs have the same amplitudes $A_k$.

942: \\

943: %

944: {\it iv)} The crude combinatorial estimates of interactions in the proof 

945: of Proposition~\ref{prop:minpunc} cannot be used to obtain corrections to the asymptotic 

946: behaviour. See also the discussion in the conclusion.

947: 

948: \subsection{Polygons with $r$ punctures of bounded size \label{sec:scaling_bound}}

949: 

950: The arguments in the above proof can be applied to 

951: obtain results for polygon models with a finite number 

952: of punctures of bounded size. 

953: The following theorem generalises Proposition~\ref{prop:minpunc} and serves as  

954: preparation for the next section, where the case of a finite number 

955: of punctures of arbitrary size is discussed.

956: For a class of punctured self-avoiding polygons, consider 

957: their area moment coefficients

958: %

959: \begin{displaymath}

960: p_m^{(r,k,s)}:= \sum_n n^kp_{m,n}^{(r,s)},

961: \end{displaymath}

962: %

963: where $p_{m,n}^{(r,s)}$ denotes the number of polygons in the 

964: class of half-perimeter $m$ and area $n$ with $r$ punctures,

965:  $r\in\mathbb N_0$, obeying the condition that the sum of the 

966: half-perimeter values of the puncturing polygons equals $s$. 

967: For simplicity of notation, we write $p_{m,n}:=p_{m,n}^{(0,0)}$, 

968: $p_m^{(k)}:=p_{m}^{(0,k,0)}$ and $p_m:=p_m^{(0)}$.

969: 

970: \begin{theorem}\label{theo:bounded}

971: Assume that, for a class of self-avoiding polygons without 

972: punctures, the area moment coefficients $p_m^{(k)}$ have 

973: the asymptotic form, for $k\in\mathbb N_0$,

974: %

975: \begin{displaymath}

976: p_m^{(k)} \sim A_k x_c^{-m} m^{\gamma_k-1} \qquad 

977: (m\to\infty),

978: \end{displaymath} 

979: %

980: for numbers $A_k>0$,  $x_c>0$ and $\gamma_k=(k-\theta)/\phi$, 

981: where $\theta$ and $\phi$ are real constants and 

982: $0<\phi<1$. Denote its half-perimeter generating function 

983: by ${\cal P}(x)=\left(\sum_{m\ge0} x^m\, p_m\right)$. 

984: Fix $r\ge1$ und $s\in\mathbb N$ such that $[x^s]({\cal P}(x))^r\ne0$. 

985: Then, the area moment coefficient $p_m^{(r,k,s)}$ of the polygon class 

986: with $r\ge1$ punctures whose half-perimeter sum equals is

987: asymptotically given by, for $k\in\mathbb N_0$,

988: %

989: \begin{equation}\label{eq:Aks}

990: p_m^{(r,k,s)} \sim A_k^{(r,s)} x_c^{-m}m^{\gamma_k^{(r)}-1}

991: \qquad (m\to\infty),

992: \end{equation}

993: %

994: where $\gamma_k^{(r)}=\gamma_{k+r}$ and $A_{k}^{(r,s)}=\frac{A_{k+r}}{r!}

995: x_c^s[x^s]({\cal P}(x))^r$.

996: \end{theorem}

997: 

998: 

999: \noindent {\bf Remarks.} {\it i)} 

1000: With $s=2$, Theorem~\ref{theo:bounded} reduces 

1001: to Proposition~\ref{prop:minpunc}.

1002: By summation, we also obtain the asymptotic behaviour for models with 

1003: $r$ punctures of total half-perimeter less or equal to $s$. Note that

1004: we have the formal identity 

1005: %

1006: \begin{displaymath}

1007: \sum_{s=0}^\infty x^s[x^s]({\cal P}(x))^r= ({\cal P}(x))^r.

1008: \end{displaymath}

1009: %

1010: The above expressions are convergent for $|x|<x_c$. If $\theta>0$, the sum is also

1011: convergent in the limit $x\nearrow x_c$.\\

1012: {\it ii)} The remarks following the proof 

1013: of Proposition~\ref{prop:minpunc} also apply to Theorem~\ref{theo:bounded}.

1014: 

1015: \begin{proof}

1016: This proof is a direct extension of the proof of Proposition~\ref{prop:minpunc} to the 

1017: case of a finite number of punctures of bounded size. We consider a model of 

1018: punctured polygons where, for fixed $s$, the 

1019: $r$ punctures of half-perimeter $s_i$ and area $t_i$ satisfy 

1020: $s_1+\ldots+s_r= s$. 

1021: We give an asymptotic estimate for $p_m^{(r,k,s)}$. Let a polygon 

1022: $\Pol$ of half-perimeter $m-|\boldsymbol s|$ and of area $n+

1023: |\boldsymbol t|$, where $|\boldsymbol s|=s_1+\ldots+s_r$ and 

1024: $|\boldsymbol t|=t_1+\ldots+t_r$, be given. To obtain an upper 

1025: bound for $p_m^{(r,k,s)}$, ignore all interactions between 

1026: components of a punctured polygon. Recall that two polygons 

1027: interact if their boundary curves have non-empty intersection. 

1028: The number of ways of placing $r$ punctures inside $\Pol$ is 

1029: clearly smaller than 

1030: %

1031: \begin{displaymath}

1032: (n+|\boldsymbol t|)^r/r!.

1033: \end{displaymath}

1034: % 

1035: This bound is obtained by considering the number of ways of placing 

1036: the lower left corner of each puncture on each square plaquette 

1037: inside the polygon. Note that, unlike in the proof of Proposition 

1038: \ref{prop:minpunc}, this bound also counts configurations where punctures 

1039: protrude from the boundary of $\Pol$. We will compensate for these 

1040: over-counted configurations when deriving a lower bound for 

1041: $p_m^{(r,k,s)}$. We have

1042: %

1043: \begin{eqnarray}

1044: p_m^{(r,k,s)} &\le {\widetilde p}_m^{(r,k,s)}:= \frac{1}{r!}

1045: \sum_{|\boldsymbol s|= s}\sum_{t_i}\sum_{n\ge 1} n^k (n+|\boldsymbol t|)^rp_{m-

1046: |\boldsymbol s|,n+|\boldsymbol t|}\prod_{i=1}^r p_{s_i,t_i}\nonumber\\

1047: &=\frac{1}{r!} \sum_{|\boldsymbol s|= s}\sum_{t_i}\sum_{n\ge |\boldsymbol t|+1}

1048: (n-|\boldsymbol t|)^k n^r p_{m-|\boldsymbol s|,n}\prod_{i=1}^r \label{form:theo1eq}

1049: p_{s_i,t_i},

1050: \end{eqnarray}

1051: %

1052: where the first sum is over the variables $s_1,\ldots,s_r$ 

1053: subject to the restriction $|\boldsymbol s|= s$, and the second

1054: sum is over all values of the variables $t_1,\ldots,t_r$.

1055: Note that, for $m$ fixed, all sums are finite. 

1056: Invoking Bernoulli's inequality, we obtain the bound

1057: %

1058: \begin{eqnarray*}

1059: \frac{1}{r!} \sum_{|\boldsymbol s|= s}\sum_{t_i}&\sum_{n\ge |\boldsymbol t|+1} 

1060: \left( n^{k+r}-k|\boldsymbol t|\, n^{k+r-1}\right) 

1061: p_{m-|\boldsymbol s|,n}\prod_{i=1}^r p_{s_i,t_i}\le 

1062: {\widetilde p}_m^{(r,k,s)}\\

1063: &\le \frac{1}{r!}\sum_{|\boldsymbol s|= s}\sum_{t_i}\sum_{n\ge |\boldsymbol t|+1}

1064: n^{k+r}p_{m-|\boldsymbol s|,n}\prod_{i=1}^r p_{s_i,t_i}.

1065: \end{eqnarray*}

1066: %

1067: Consider first the asymptotic behaviour of the expression

1068: %

1069: \begin{displaymath}

1070: {\widetilde a}_{m,s}:= \sum_{|\boldsymbol s|= s}\sum_{t_i} 

1071: \sum_{n\ge |\boldsymbol t|+1} n^{k+r}p_{m-|\boldsymbol s|,n}

1072: \prod_{i=1}^r p_{s_i,t_i}.

1073: \end{displaymath}

1074: % 

1075: If $m\ge |\boldsymbol s|^2+|\boldsymbol s|+2$, then the

1076: lower bound of summation on the index $n$ may be replaced by zero.

1077: This follows from the estimate $t_i\le s_i^2$, being valid

1078: for every self-avoiding polygon of half-perimeter $s_i$ and area $t_i$.

1079: Thus $|\boldsymbol t|\le |\boldsymbol s|^2$, and we argue

1080: that $n\ge m-|\boldsymbol s|-1\ge |\boldsymbol s|^2+1\ge 

1081: |\boldsymbol t|+1$. We thus get for $m$ sufficiently large 

1082: %

1083: \begin{eqnarray*}

1084: {\widetilde a}_{m,s} = \sum_{|\boldsymbol s|= s} 

1085: p_{m-|\boldsymbol s|}^{(k+r)} \prod_{i=1}^r p_{s_i}

1086: \sim p_m^{(k+r)} \left( \sum_{|\boldsymbol s|= s} 

1087: x_c^{|\boldsymbol s|}\prod_{i=1}^r p_{s_i}\right) \qquad (m\to\infty),

1088: \end{eqnarray*}

1089: %

1090: where the sum in brackets is finite.

1091: We now analyse the second term in the estimate derived from the

1092: Bernoulli inequality. To this end, define

1093: %

1094: \begin{displaymath}

1095: {\widetilde b}_{m,s}:= \sum_{|\boldsymbol s|= s}\sum_{t_i} 

1096: \sum_{n\ge |\boldsymbol t|+1} |\boldsymbol t|n^{k+r-1}

1097: p_{m-|\boldsymbol s|,n}\prod_{i=1}^r p_{s_i,t_i}.

1098: \end{displaymath}

1099: %

1100: Using the estimate $|\boldsymbol t|\le |\boldsymbol s|^2$, we get

1101: %

1102: \begin{eqnarray*}

1103: {\widetilde b}_{m,s} \le s^2 \left(\sum_{|\boldsymbol s|= s} 

1104: p_{m-|\boldsymbol s|}^{(k+r-1)} \prod_{i=1}^r p_{s_i}\right)

1105: \sim s^2  p_m^{(k+r-1)} \left( \sum_{|\boldsymbol s|= s} 

1106: x_c^{|\boldsymbol s|}\prod_{i=1}^r p_{s_i}\right) \qquad (m\to\infty).

1107: \end{eqnarray*}

1108: %

1109: Now set  $b_{m,s}:={\widetilde b}_{m,s}/(x_c^{-m}m^{\gamma_{k}^{(r)}-1})$.

1110: The above estimate yields $\lim_{m\to\infty} b_{m,s}=0$, since $0<\phi<1$.

1111: 

1112: We now derive a lower bound for $p_m^{(r,k,s)}$ by subtracting

1113: from ${\widetilde p}_m^{(r,k,s)}$ an upper bound on the contributions 

1114: arising from puncture-puncture interactions and from puncture-boundary 

1115: interactions. We will show that the lower bound coincides asymptotically

1116: with the upper bound, which then implies the assertion of the theorem

1117: %

1118: \begin{displaymath}

1119: p_m^{(k,r,s)}\sim \frac{A_{k+r}}{r!} 

1120: \left( \sum_{|\boldsymbol s|= s} x_c^{|\boldsymbol s|}

1121: \prod_{i=1}^r p_{s_i}\right) x_c^{-m}m^{\gamma_{k+r}-1}

1122: \qquad (m\to\infty).

1123: \end{displaymath}

1124: %

1125: For any polygon $\Pol$, the number of puncture-puncture interactions 

1126: between $r>1$ punctures is smaller than the number of 

1127: puncture-puncture interactions of two punctures with the remaining

1128: $r-2$ punctures occuring at arbitrary positions in the polygon.

1129: We thus get the upper bound

1130: %

1131: \begin{displaymath}

1132: (t_1+4s_1)t_2(n+|\boldsymbol t|)(n+|\boldsymbol t|)^{r-2}

1133: \le 6 t_1t_2 (n+|\boldsymbol t|)^{r-1},

1134: \end{displaymath}

1135: where we used $t_1\ge s_1-1$. The factor $(t_1+4s_1)t_2$ 

1136: bounds the number of configurations of two interacting punctures,

1137: and the factor $(n+|\boldsymbol t|)^{r-2}$ arises from allowing

1138: arbitrary positions of the remaining $r-2$ punctures. Define

1139: %

1140: \begin{displaymath}

1141: {\widetilde c}_{m,s}:= \sum_{|\boldsymbol s|= s}\sum_{t_i} 

1142: \sum_{n} t_1t_2 n^k (n+|\boldsymbol t|)^{r-1}

1143: p_{m-|\boldsymbol s|,n+|\boldsymbol t|}\prod_{i=1}^r p_{s_i,t_i}

1144: \le s^4 \left(\sum_{|\boldsymbol s|= s} 

1145: p_{m-|\boldsymbol s|}^{(k+r-1)}\prod_{i=1}^r p_{s_i}\right),

1146: \end{displaymath}

1147: %

1148: where we used $t_i\le|\boldsymbol t|\le|\boldsymbol s|^2$ for the

1149: last inequality. Setting $c_{m,s}:={\widetilde c}_{m,s}/

1150: (x_c^{-m}m^{\gamma_{k}^{(r)}-1})$, we infer that $\lim_{m\to\infty}

1151: c_{m,s}=0$. We have shown that for fixed $s$ the 

1152: puncture-puncture interactions are asymptotically irrelevant.

1153: 

1154: We finally estimate the puncture-boundary interactions. This is done

1155: similarly to the above treatment of puncture-puncture interactions.

1156: The number of puncture-boundary interactions is bounded from above by

1157: %

1158: \begin{displaymath}

1159: \sum_{j=1}^r4^j(m-|\boldsymbol s|)^j s_1\cdot\ldots\cdot s_j \,

1160: (n+|\boldsymbol t|)^{r-j},

1161: \end{displaymath}

1162: %

1163: where $j$ punctures interact with the boundary, each contributing

1164: a factor $4(m-|\boldsymbol s|)s_i$, and $r-j$ punctures have arbitrary

1165: positions, each contributing a factor $(n+|\boldsymbol t|)$. Note that

1166: the over-counted configurations in ${\widetilde p}_m^{(r,k,s)}$, which 

1167: protrude from the boundary, are compensated for by the above estimate.

1168: Define

1169: %

1170: \begin{eqnarray*}

1171: {\widetilde d}_{m,s}&:= \sum_{|\boldsymbol s|= s}\sum_{t_i} 

1172: \sum_{n} (m-|\boldsymbol s|)^j n^k (n+|\boldsymbol t|)^{r-j}s_1\cdot

1173: \ldots\cdot s_j\,p_{m-|\boldsymbol s|,n+|\boldsymbol t|}\prod_{i=1}^r 

1174: p_{s_i,t_i}\\

1175: &\le (m-s)^j s^{j} \left(\sum_{|\boldsymbol s|= s} 

1176: p_{m-|\boldsymbol s|}^{(k+r-j)}\prod_{i=1}^r p_{s_i}\right).

1177: \end{eqnarray*}

1178: %

1179: Defining $d_{m,s}:={\widetilde d}_{m,s}/

1180: (x_c^{-m}m^{\gamma_{k}^{(r)}-1})$, we infer that $\lim_{m\to\infty}

1181: d_{m,s}=0$. We have shown that for fixed $s$ the puncture-boundary 

1182: interactions are asymptotically irrelevant. This completes  the proof. 

1183: \end{proof}

1184: 

1185: 

1186: \subsection{Polygons with $r$ punctures of arbitrary size 

1187: \label{sec:scaling_arbitrary}}

1188: 

1189: For a class of punctured self-avoiding polygons, consider 

1190: for $k\in\mathbb N_0$ their area moment coefficients

1191: %

1192: \begin{displaymath}

1193: p_m^{(r,k)}:= \sum_n n^kp_{m,n}^{(r)},

1194: \end{displaymath}

1195: %

1196: where $p_{m,n}^{(r)}:=\sum_{s=0}^\infty p_{m,n}^{(r,s)}<\infty$ denotes 

1197: the number of polygons in the class of half-perimeter $m$ and area $n$ with 

1198: $r$ punctures of arbitrary size, $r\in\mathbb N_0$. For simplicity 

1199: of notation, we write $p_{m,n}=p_{m,n}^{(0)}$ and $p_m^{(k)}=

1200: p_{m}^{(0,k)}$. In the sequel, we will use the area moment generating 

1201: functions $\PGf_k(x)=\sum p_m^{(k)}x^m$ of the model without punctures.

1202: 

1203: \begin{theorem}\label{theo:arb}

1204: Assume that, for a class of self-avoiding polygons without 

1205: punctures, the area moment coefficients $p_m^{(k)}$ 

1206: have the asymptotic form, for $k\in\mathbb N_0$,

1207: %

1208: \begin{displaymath}

1209: p_m^{(k)} \sim A_k x_c^{-m} m^{\gamma_k-1} \qquad 

1210: (m\to\infty)

1211: \end{displaymath} 

1212: %

1213: for numbers $A_k>0$,  $x_c>0$ and $\gamma_k=(k-\theta)/\phi$, 

1214: where $0<\phi<1$. Let $\PGf_k(x)=\sum p_m^{(k)}x^m$ denote the $k^{th}$ area 

1215: moment generating function. 

1216: 

1217: Then, the area moment coefficient $p_m^{(r,k)}$ of the polygon class with $r\ge1$ 

1218: punctures of arbitrary size is, for $k\in\mathbb N_0$, bounded from above by

1219: %

1220: \begin{displaymath}

1221: p_m^{(r,k)} \le \frac{[x^m]\PGf_{k+r}(x)(\PGf_0(x))^r}{r!}.

1222: \end{displaymath}

1223: %

1224: For finite critical perimeter generating functions, characterised by $\theta>0$, 

1225: $p_m^{(r,k)}$ is asymptotically given by, for $k\in\mathbb N_0$, 

1226: %

1227: \begin{equation}\label{form:t2}

1228: p_m^{(r,k)} \sim \frac{[x^m]\PGf_{k+r}(x)(\PGf_0(x))^r}{r!} \sim

1229: A_k^{(r)}x_c^{-m} m^{\gamma_{k+r}-1}

1230: \qquad (m\to\infty),

1231: \end{equation}

1232: %

1233: where the amplitudes $A_k^{(r)}$ are given by

1234: %

1235: \begin{equation}\label{eq:Aka}

1236: A_k^{(r)}=\frac{A_{k+r}(\PGf_0(x_c))^r}{r!},

1237: \end{equation}

1238: %

1239: where $\PGf_0(x_c):=\lim_{x\nearrow x_c}\PGf_0(x)<\infty$ is the critical amplitude of the

1240: half-perimeter generating function.

1241: \end{theorem}

1242: 

1243: 

1244: \noindent {\bf Remarks.} {\it i)}

1245: The asymptotic form Eq.~(\ref{form:t2}) is formally obtained from 

1246: Theorem~\ref{theo:bounded} in the limit of infinite puncture size, 

1247: see Remark {\it i)} after Theorem~\ref{theo:bounded}. This observation 

1248: is also the main ingredient of the following proof, by noting that 

1249: the upper bound has the same asymptotic behaviour.

1250: \\

1251: {\it ii)} For staircase polygons, where 

1252: $\theta=1/3$ and $\phi=2/3$, the assumptions of Theorem 

1253: \ref{theo:arb} are satisfied. For self-avoiding 

1254: polygons, we have the numerically very well established

1255: values $\theta=1$ and $\phi=2/3$, which we believe to describe

1256: the asymptotic behaviour of SAPs.

1257: For models satisfying $\theta<0$, the upper bound generally does

1258: not coincide asymptotically with $p_m^{(r,k)}$. An example

1259: of failure is rectangles with a single puncture.

1260: 

1261: \begin{proof}

1262: We obtain as in the proof of Theorem~\ref{theo:bounded} an upper bound 

1263: ${\widetilde p}_m^{(r,k)}$ for the area moment 

1264: coefficients $p_m^{(r,k)}$. It is given by

1265: %

1266: \begin{displaymath}

1267: p_m^{(r,k)}\le{\widetilde p}_m^{(r,k)} :=\frac{1}{r!}\sum_{s=0}^m 

1268: \sum_{|{\bf s}|=s}p_{m-|{\bf s}|}^{(k+r)}\prod_{i=1}^r p_{s_i} 

1269: =\frac{1}{r!}[x^m] \PGf_{k+r}(x) (\PGf_0(x))^r.

1270: \end{displaymath}

1271: %

1272: Assume in the following that $\theta>0$. The asymptotic behaviour 

1273: of the rhs of (\ref{form:t2}) follows by $r$-fold application of 

1274: Lemma 1, which is given in the appendix.

1275: Note that, for $M$ arbitrary, we have by definition 

1276: %

1277: \begin{displaymath}

1278: p_m^{(r,k)}\ge\sum_{s=0}^M p_m^{(r,k,s)},

1279: \end{displaymath}

1280: %

1281: where $p_m^{(r,k,s)}$ is the number of $r$-punctured polygons,

1282: whose punctures have total perimeter equal to $s$.

1283: Theorem~\ref{theo:bounded} implies that the above sum is,

1284: for $M$ sufficiently large, asymptotically in $m$, arbitrarily close

1285: to the upper bound ${\widetilde p}_m^{(r,k)}$. See also the 

1286: remark following Theorem~\ref{theo:bounded}.  This 

1287: yields the statement of the theorem.

1288: \end{proof}

1289: 

1290: \subsection{Limit distribution of area and scaling 

1291: function conjectures \label{sec:scaling_func}}

1292: 

1293: We first discuss the implications of the previous

1294: results on the asymptotic area law of polygon models

1295: with punctures. By an application of Proposition 

1296: \ref{theo:general}, Theorem \ref{theo:bounded} and 

1297: Theorem \ref{theo:arb} immediately yield the following 

1298: result:

1299: 

1300: \begin{theorem}\label{thm:limdist}

1301: Assume that, for a class of self-avoiding polygons without 

1302: punctures, the area moment coefficients $p_m^{(k)}$ 

1303: have the asymptotic form, for $k\in\mathbb N_0$,

1304: %

1305: \begin{displaymath}

1306: p_m^{(k)} \sim A_k x_c^{-m} m^{\gamma_k-1} \qquad 

1307: (m\to\infty)

1308: \end{displaymath} 

1309: %

1310: for numbers $A_k>0$,  $x_c>0$ and $\gamma_k=(k-\theta)/\phi$, where $0<\phi<1$.

1311: Assume further that the numbers $A_k$ satisfy the Carleman condition

1312: %

1313: \begin{displaymath}

1314: \sum_{k\ge0} (A_{2k})^{-1/{2k}}=+\infty.

1315: \end{displaymath}

1316: %

1317: Denote the half-perimeter generating function of the model

1318: by ${\cal P}(x)=\left(\sum_{m\ge0} x^m\, p_m\right)$. 

1319: \begin{itemize}

1320: \item[\it i)] 

1321: Consider for $r\ge1$ the corresponding model with $r$ 

1322: punctures of bounded size, whose half-perimeter sum equals $s\in \mathbb N$, such

1323: that $[x^s]({\cal P}(x))^r\ne0$.

1324: Denote the random variables of area in the uniform fixed 

1325: perimeter ensemble by $\widetilde X_m^{(r,s)}$. Then,

1326: we have convergence in distribution,

1327: %

1328: \begin{displaymath}

1329: \frac{\widetilde X_m^{(r,s)}}{m^{1/\phi}} \stackrel{d}{\longrightarrow} X^{(r,s)} 

1330: \qquad (m\to\infty),

1331: \end{displaymath}

1332: %

1333: for a uniquely defined random variable $X^{(r,s)}$ with moments 

1334: %

1335: \begin{displaymath}

1336: \mathbb E[(X^{(r,s)})^k]=\frac{A_k^{(r,s)}}{A_r},

1337: \end{displaymath}

1338: %

1339: where the numbers $A_k^{(r,s)}$ are those of Theorem~\ref{theo:bounded}.

1340: We also have moment convergence.

1341: \item[\it ii)]

1342: Let $\widetilde X_m^{(r)}$ denote the

1343: random variable of area in the fixed perimeter ensemble for

1344: the model with $r\ge1$ punctures of unbounded size. If $\theta>0$, then

1345: we have convergence in distribution,

1346: %

1347: \begin{displaymath}

1348: \frac{\widetilde X_m^{(r)}}{m^{1/\phi}} \stackrel{d}{\longrightarrow} X^{(r)} \qquad (m\to\infty)

1349: \end{displaymath}

1350: %

1351: for a uniquely defined random variable $X^{(r)}$ with moments 

1352: %

1353: \begin{displaymath} 

1354: \mathbb E[(X^{(r)})^k]=\frac{A_k^{(r)}}{A_r},

1355: \end{displaymath}

1356: %

1357: where the numbers $A_k^{(r)}$ are those of Theorem~\ref{theo:arb}.

1358: We also have moment convergence.

1359: \item[\it iii)]

1360: If $\theta>0$, the random variables $X^{(r)}$ and $X^{(r,s)}$ are related by

1361: %

1362: \begin{displaymath}

1363: x_c^s[x^s]({\cal P}(x))^r\, X^{(r)}=({\cal P}(x_c))^r\, X^{(r,s)},

1364: \end{displaymath}

1365: %

1366: where ${\cal P}(x)$ is the half-perimeter generating function

1367: of the polygon model without punctures, and where $\mathcal{P}(x_c)=

1368: \lim_{x\nearrow  x_c}\mathcal{P}(x)<\infty$.

1369: %

1370: \end{itemize}

1371: \qed

1372: \end{theorem}

1373: 

1374: \noindent {\bf Remarks.} {\it i)}

1375: For a given polygon model satisfying the assumptions of 

1376: Theorem~\ref{thm:limdist}, the area moments satisfy asymptotically

1377: %

1378: \begin{displaymath}

1379: \frac{\mathbb E[(\widetilde X_m^{(r)})^k]}{k!}\sim D_k^{(r)} m^{k/\phi} 

1380: \qquad (m\to\infty),

1381: \end{displaymath}

1382: %

1383: for positive numbers $ D_k^{(r)}$.

1384: For classes of polygon models whose generating function satisfies a 

1385: $q$-functional equation with a square root as the dominant singularity

1386: of their perimeter generating function, the amplitude ratios $D_k^{(r)}/\left 

1387: [ D_1^{(r)} \right ]^k$ are universal, i.e., independent of the constants 

1388: $f_0$, $f_1$ and $x_c$, which characterise the underlying model \cite{R02,R05,R06}. This 

1389: follows from Eqs.~(\ref{eq:relation}) and (\ref{ck}) by a straightforward calculation. The 

1390: numbers are listed in Table~\ref{tab:ratios} for small values of $r$. Note that

1391: the same numbers appear for punctures of bounded size. 

1392: %

1393: \begin{table}

1394: \begin{center}

1395: \begin{tabular}{cccc}

1396: \hline \hline

1397: Amplitude & $r=0$ & $r=1$ & $r=2$ \\

1398: \hline

1399: $D_2/D_1^2$ &

1400: 0.530518$\times 10^{-0}$ & 

1401: 0.530143$\times 10^{-0}$ &

1402: 0.529356$\times 10^{-0}$

1403: \\

1404: $D_3/D_1^3$ & 

1405: 0.198944$\times 10^{-0}$ &

1406: 0.198369$\times 10^{-0}$ &

1407: 0.197361$\times 10^{-0}$

1408: \\

1409: $D_4/D_1^4$ & 

1410: 0.592379$\times 10^{-1}$& 

1411: 0.588127$\times 10^{-1}$&

1412: 0.581533$\times 10^{-1}$

1413: \\

1414: $D_5/D_1^5$ & 

1415: 0.149079$\times 10^{-1}$ &  

1416: 0.146994$\times 10^{-1}$ &

1417: 0.144042$\times 10^{-1}$

1418: \\

1419: $D_6/D_1^6$ &

1420: 0.329453$\times 10^{-2}$&

1421: 0.321705$\times 10^{-2}$&

1422: 0.311511$\times 10^{-2}$ 

1423: \\

1424: $D_7/D_1^7$ &

1425: 0.655743$\times 10^{-3}$ & 

1426: 0.632288$\times 10^{-3}$ &

1427: 0.603260$\times 10^{-3}$

1428: \\

1429: $D_8/D_1^8$ &

1430: 0.119654$\times 10^{-3}$& 

1431: 0.113600$\times 10^{-3}$&

1432: 0.106501$\times 10^{-3}$

1433: \\

1434: $D_9/D_1^9$ & 

1435: 0.202754$\times 10^{-4}$ & 

1436: 0.189015$\times 10^{-4}$ &

1437: 0.173673$\times 10^{-4}$

1438:  \\

1439: $D_{10}/D_1^{10}$ &

1440: 0.322150$\times 10^{-5}$ & 

1441: 0.294132$\times 10^{-5}$ &

1442: 0.264251$\times 10^{-5}$

1443: \\

1444: \hline \hline

1445: \end{tabular}

1446: \end{center}

1447: \caption{\label{tab:ratios} 

1448: Universal amplitude ratios for staircase polygons with $r$ punctures.}

1449: \end{table} 

1450: %

1451: \\

1452: {\it ii)}

1453: For the above class of models, explicit expressions for the 

1454: asymptotic behaviour of their moment generating functions and 

1455: their probability distributions can be derived from the area 

1456: amplitude series via inverse Laplace transform techniques. 

1457: Since the resulting expressions are quite cumbersome, we do not 

1458: give them here. For $r=0$ the corresponding limit distribution 

1459: of area is the Airy distribution \cite{L84,T91,FL01}. The 

1460: extension to $r\ge1$ is straightforward.

1461: As mentioned above, for $r=0$ the amplitude ratios are found 

1462: to coincide with those of (rooted) self-avoiding polygons to a 

1463: high degree of numerical accuracy \cite{RJG03}. If the conjecture

1464: holds true that they agree {\it exactly}, then the above expressions 

1465: for limit distributions also appear for rooted self-avoiding polygons, 

1466: for all values of $r$. See Section 6 for a detailed numerical 

1467: analysis.

1468: 

1469: \medskip

1470: 

1471: We now discuss the relations between the area amplitude series

1472: $F(z)$ of the polygon model without punctures and $F^{(r)}(z)$ of

1473: the polygon model with $r$ punctures. Since all of our models have an

1474: entire moment generating function, the Carleman condition is satisfied,

1475: and Theorem \ref{theo:bounded} and Theorem \ref{theo:arb} yield, by a 

1476: straightforward calculation, the following result.

1477: 

1478: \begin{theorem}\label{thm:amplfct}

1479: Assume that, for a class of self-avoiding polygons, the polygon model without punctures 

1480: has an area amplitude series, given by

1481: %

1482: \begin{displaymath}

1483: F(z)=\sum_{k\ge0} f_k z^{-\gamma_k},

1484: \end{displaymath}

1485: %

1486: where $\gamma_k=(k-\theta)/\phi\in(-1,\infty)\setminus\{0\}$ and $0<\phi<1$, and where

1487: the numbers $f_k\ne0$ are related to the amplitudes $A_k$ in Proposition 

1488: \ref{theo:general} via Eq.~(\ref{eq:relation}). Denote the half-perimeter 

1489: generating function of the model

1490: by ${\cal P}(x)=\sum_{m\ge0} x^m\, p_m$. 

1491: %

1492: \begin{itemize}

1493: \item[\it i)]

1494: Assume that $r\ge1$ and $s\in\mathbb N$ are given such that

1495: $[x^s]({\cal P}(x))^r\ne0$.

1496: Then, the corresponding model of punctured polygons with $r$ punctures 

1497: of bounded size $s$ has an area amplitude series, given by

1498: %

1499: \begin{displaymath}

1500: F^{(r,s)}(z)=\sum_{k\ge0} f_k^{(r)} z^{-\gamma_k^{(r)}},

1501: \end{displaymath}

1502: %

1503: where $\gamma_k^{(r)}=(k-\theta_r)/\phi_r$. We have 

1504: $\theta_r=\theta-r$, $\phi_r=\phi$, and

1505: %

1506: \begin{equation}\label{form:Fr}

1507: F^{(r,s)}(z)=\frac{(-1)^r}{r!}x_c^s[x^s]({\cal P}(x))^r

1508: \sum_{k\ge r} (k)_r f_k z^{-\gamma_k}.

1509: \end{equation}

1510: %

1511: \item[\it ii)]

1512: If $\theta>0$, the corresponding model of punctured polygons with $r\ge1$ punctures 

1513: of arbitrary size has an area amplitude series, given by

1514: %

1515: \begin{displaymath}

1516: F^{(r)}(z)=\frac{(-1)^r}{r!}({\cal P}(x_c))^r

1517: \sum_{k\ge r} (k)_r f_k z^{-\gamma_k},

1518: \end{displaymath}

1519: %

1520: where $\mathcal{P}(x_c)=\lim_{x\nearrow  x_c}\mathcal{P}(x)<\infty$.

1521: \end{itemize}

1522: %

1523: \qed

1524: \end{theorem}

1525: 

1526: \noindent {\bf Remarks.} {\it i)}

1527: Eq.~(\ref{form:Fr}) allows one to derive explicit expressions for the

1528: area amplitude series in terms of $F(z)$. For models where $\theta=1/3$

1529: and $\phi=2/3$ such as staircase polygons,

1530: the area amplitude series $F(z)$ satisfies the Riccati equation

1531: %

1532: \begin{equation}\label{eq:Ricc}

1533: F(z)^2-4f_1F'(z)-f_0^2z=0.

1534: \end{equation}

1535: %

1536: This can be used to show that $F^{(r,s)}(z)$ 

1537: (and also $F^{(r)}(z)$) is of the form

1538: %

1539: \begin{displaymath}

1540: F^{(r,s)}(z)=\sum_{k=0}^{r+1} p_{k,r}(z)F(z)^{k},

1541: \end{displaymath}

1542: %

1543: where $p_{k,r}(z)$ are polynomials in $z$ of degree not exceeding 

1544: $\lceil 3r/2\rceil$, and $p_{r+1,r}(z)$ is not identically vanishing. 

1545: Simple expressions for the polynomials

1546: $p_{k,r}(z)$ are not apparent.  We note, however, that such

1547: expressions appear as correction-to-scaling functions

1548: of the underlying polygon models without punctures \cite{R02}.\\

1549: {\it ii)}

1550: The model of rooted self-avoiding polygons has been found numerically

1551: to have the same type of area amplitude series as staircase polygons

1552: (with different constants $f_0$ and $f_1$). Similar considerations apply

1553: to the model of unrooted self-avoiding polygons

1554: starting from Eq.~(\ref{form:unroot}).

1555: 

1556: \medskip

1557: 

1558: We finally discuss the scaling function conjectures implied

1559: by the results of the previous subsections. For staircase polygons,

1560: the area amplitude

1561: function satisfies the differential equation~(\ref{eq:Ricc}).

1562: This differential equation has a unique solution ${\cal F}(z)$

1563: analytic for $\Re(z)\ge0$, having $F(z)$ as an asymptotic

1564: expansion at infinity, ${\cal F}(z)\sim F(z)$ as $z\to\infty$. 

1565: The function ${\cal F}(z)$ is explicitly given by

1566: %

1567: \begin{displaymath}

1568: {\cal F}(z)=-4f_1 \frac{{\rm d}}{{\rm d}z}\log\mbox{Ai}

1569: \left(\left(\frac{f_0}{4f_1}\right)^{2/3}z\right),

1570: \end{displaymath}

1571: %

1572: and this function coincides with the scaling function of

1573: staircase polygons Eq.~(\ref{sfstair}). 

1574: 

1575: In analogy to the above observation, we conjecture that

1576: the area amplitude series for punctured staircase

1577: polygons determine their scaling functions. Likewise, we conjecture that

1578: the area amplitude series for punctured rooted self-avoiding

1579: polygons determine their scaling functions.

1580: 

1581: \begin{conjecture}\label{con1}

1582: Let $r\ge1$ and $s\ge2$ be given. For staircase polygons and rooted self-avoiding polygons, 

1583: the area amplitude series $F^{(r,s)}(z)$ and $F^{(r)}(z)$ of

1584: Theorem~\ref{thm:amplfct} uniquely define functions ${\cal F}^{(r,s)}(z)$ 

1585: and ${\cal F}^{(r)}(z)$ analytic for $\Re(z)> z_0$ and non-analytic at $z=z_0$, 

1586: for some negative real number $z_0<0$. We conjecture that the functions 

1587: ${\cal F}^{(r,s)}(z):(z_0,\infty)\to\mathbb R$ and 

1588: ${\cal F}^{(r)}(z):(z_0,\infty)\to\mathbb R$ are scaling functions

1589: as in  Definition~\ref{def:sf},

1590: %

1591: \begin{eqnarray*}

1592: {\cal P}^{(r,s)}(x,q) &\sim  (1-q)^{1/3-r}{\cal F}^{(r,s)}

1593: \left(\frac{x_c-x}{(1-q)^{2/3}}\right)\\

1594: {\cal P}^{(r)}(x,q) &\sim  (1-q)^{1/3-r}{\cal F}^{(r)}

1595: \left(\frac{x_c-x}{(1-q)^{2/3}}\right).

1596: \end{eqnarray*}

1597: %

1598: \end{conjecture}

1599: 

1600: \medskip

1601: 

1602: \noindent {\bf Remark.}

1603: The above conjecture has the following implications.\\

1604: {\it i)} Staircase polygons with a single minimal 

1605: puncture  specialise to Eq.~(\ref{stsf}).\\

1606: {\it ii)} Up to constant factors, the scaling 

1607: form of the model with $r$ punctures is obtained as the $r^{th}$ 

1608: derivative w.r.t.~$q$ of the scaling form of the model without 

1609: punctures, as can be proved by induction. As derivatives

1610: can be interpreted combinatorially as marking, this reflects the fact 

1611: underlying the proofs in this section that punctures may be regarded 

1612: as being asymptotically independent, and boundary effects do not 

1613: play a role asymptotically.\\

1614: {\it iii)}

1615: Ignoring questions of analyticity, a (formal) calculation yields 

1616: that the functions ${\cal F}^{(r)}$ (and ${\cal F}^{(r,s)}$) lead, 

1617: for both staircase polygons and (unrooted) self-avoiding polygons, 

1618: to the same critical exponents in the branched polymer phase as 

1619: those conjectured previously \cite{JvR92,GJWE00}. These are obtained from 

1620: the singular behaviour of ${\cal F}^{(r)}$ about the singularity of 

1621: smallest modulus on the negative axis, i.e., at the first zero of 

1622: the Airy-function on the negative axis, see \cite[Sec~1]{R02}. The 

1623: fact that $\PGf^{(r)}(x,q)$ is obtained from $\PGf(x,q)$ by $r$-fold 

1624: differentiation yields the result.

1625: 

1626: 

1627: \section{Computer enumerations \label{sec:enum}}

1628: 

1629: Here we briefly outline which algorithms were used to derive the series expansions

1630: for the area moments of punctured polygons. In most cases (SAPs and punctured

1631: staircase polygons) the algorithms are simple generalisations of previous 

1632: algorithms already described in detail in other papers, referenced below. In

1633: these cases we give brief details of the length of the series and the amount

1634: of CPU time used. Only in the case of staircase polygons with minimal punctures

1635: did we write a new specific algorithm which we shall describe in some detail.

1636: 

1637: The series for punctured self-avoiding polygons were calculated using a simple

1638: generalisation of the parallel version of the algorithm we used previously 

1639: to enumerate ordinary SAPs \cite{IJ03}. In each case (SAPs with one or two

1640: minimal punctures and SAPs with one or two arbitrary punctures) we calculated the

1641: area moments up to $k=10$ for SAPs to total perimeter 100. Since the smallest such

1642: SAPs have perimeter 16 and 24 this results in series with 42 and 38 non-zero

1643: terms, respectively. The total CPU time required was about 5000 hours for

1644: each of the once punctured SAP problems and up to 3000 hours for the twice

1645: punctured problems. The bulk of these calculations were performed on the old facility

1646: of the Australian Partnership for Advanced Computing (APAC), which

1647: was a Compaq Server Cluster with ES45 nodes with 1GHz Alpha chips (this facility

1648: has since been replaced by an SGI Altix cluster).

1649: 

1650: In \cite{GJ06b} we used a very efficient algorithm to enumerate once punctured

1651: staircase polygons. The algorithm is easily generalised in order to calculate

1652: area moments which we have done to perimeter 718 ($k=1$), 598 ($k=2$) and 

1653: 506 ($k=3$ to 10). It is also quite straight-forward to generalise the algorithm

1654: to count twice punctured staircase polygons and in this case we obtained the

1655: series to perimeter 502 for $k=0$, perimeter 450 for $k=1$ and 2 and to perimeter 

1656: 302 for $k=3$ to 10. It is also easy to extract data for staircase polygons with 

1657: punctures of fixed combined perimeter. 

1658: In each case we used around 1000 CPU hours on the APAC Altix cluster which

1659: use 1.6GHz Intel Itanium 2 chips.

1660: 

1661: 

1662: \begin{figure}

1663: \begin{center}

1664: \includegraphics[scale=0.6]{transfer}

1665: \end{center}

1666: \caption{\label{fig:TM} 

1667: Illustration of the transfer matrix boundary line and local updating rules.

1668: }

1669: \end{figure}

1670: 

1671: Finally we describe the algorithm used to enumerate minimally punctured staircase

1672: polygons. The algorithm is based on so-called transfer matrix techniques.

1673: The basic idea is to count the number of polygons by bisecting the lattice with

1674: a boundary line. In the left panel of Fig.~\ref{fig:TM} we show how such a boundary

1675: (the first medium thick line) will intersect the polygon in several places.

1676: The first and last occupied edges intersected by the boundary line are the

1677: directed walks constituting the outer staircase polygon. The other occupied

1678: edges (if any) belong to the minimal punctures. In a calculation to maximal

1679: half-perimeter  $m$ we need only consider intersections with widths up to $w=m/2$.

1680: Any intersection pattern (or signature) can be specified by a string of occupation 

1681: variables, $S = \{\sigma_0,\sigma_1,\ldots \sigma_w \}$, where $\sigma_i=0$, 1 or 2

1682: if edge number $i$ is empty, an occupied outer edge or an occupied edge

1683: part of a minimal puncture, respectively. We could use the same symbol for all

1684: occupied edges but it is convenient to explicitly distinguish between the two cases.

1685: For each signature we keep a generating function which keeps track of the number 

1686: of configurations (to the left of the boundary line), that is, it counts the number 

1687: of possible partially completed polygons with a given signature. In order

1688: to count the total number of punctured polygons we move the boundary line

1689: to the right column by column with each column built up one vertex at a time.

1690: In the left panel of  Fig.~\ref{fig:TM} we have also shown a typical move of the boundary 

1691: line, which starts in the position given by the second medium thick line and where

1692: we add two new edges to the lattice by moving the kink in the boundary line to

1693: the position given by the thin lines. As we move the boundary line to a

1694: new position we calculate the associated generating functions (the updating rules will

1695: be given below). Formally we can view this transformation between signatures as a matrix 

1696: multiplication (hence our use of the nomenclature {\em transfer matrix algorithm}).

1697: However, as can be readily seen, the transfer matrix is extremely sparse and there 

1698: is no reason to list it explicitly (it is given {\em implicitly} by the updating rules).

1699: 

1700: We start the calculation with the initial signature $\{1,1,0, \ldots ,0\}$, which

1701: corresponds to inserting the two steps of the outer walks in the lower left corner 

1702: (the count of this configuration is 1). As the boundary line is moved it passes over 

1703: a vertex and the updating depends on the states of the edges below and to the left of 

1704: this vertex. After the move we `insert' the edges to the right and above the vertex. 

1705: There  are four possible local configurations of the `incoming' edges as illustrated 

1706: in the middle panels of Fig.~\ref{fig:TM}: Both edges are empty, one of the edges is 

1707: occupied and the other edge is empty or both edges are occupied. 

1708: 

1709: 

1710: \begin{description}

1711: \item{\bf Both edges empty:} If both incoming edges are empty then both outgoing edges

1712: can be empty. Else we may insert two new steps which {\em must} be part of a minimal puncture

1713: (the outgoing edges are in state `2'). This is {\em only} possible if the vertex is in the

1714: interior of the polygon (there is an edge in state `1' both below and above the vertex).

1715: \item{\bf Left edge empty, bottom edge occupied:} The walk occupying the incoming edge must 

1716: be continued along an outgoing edge. If the occupied edge is part of the external

1717: polygon (in state `1') there are no restrictions. If the occupied edge is part of 

1718: a minimal puncture the walk can only be continued along the edge to the right of the

1719: vertex (otherwise we would not get a minimal puncture).

1720: \item{\bf Left edge occupied, bottom edge empty:} This is similar to the previous

1721: case except that an edge in state `2' must be continued along the edge above the vertex.

1722: \item{\bf Both edges occupied:} If both edges are in state `2' we close the puncture

1723: and the new edges are empty. If the incoming edges are in state `1' we have closed

1724: the outer polygon and then we add the count to the running total for the generating

1725: function.

1726: \end{description}

1727: In the last panel of Fig.~\ref{fig:TM} we show how the updating rules given above

1728: through a sequence of moves of the boundary line gives rise to a minimal puncture.

1729: 

1730: The perimeter of the completed polygon is given by the position of the boundary

1731: line when the polygon is closed, e.g., if we have taken $k$ steps in the

1732: $x$-direction (completed $k$ columns) and moved the kink $l$ steps in the $y$-direction

1733: then the outer half-perimeter is $k+l$ and the total half-perimeter in $k+l+2r$.

1734: So we need only keep track of the number of punctures $r$. This is done

1735: by associating a truncated polynomial $P_S(x)=\sum p_r x^r$ with each signature, where

1736: the coefficient $p_r$ is the number of partially completed polygons with $r$

1737: punctures of the signature $S$. As a new signature $S'$ is created from $S$

1738: we set $P_{S'}(x)=P_{S'}(x)+x^{\delta}P_S(x)$, where $\delta =1$ if an additional

1739: puncture is inserted (as in the first case described above) or 0 otherwise.

1740: The extension to the calculation of area moments is described in \cite{IJ00}.

1741: 

1742: \section{Numerical analysis \label{sec:numeric}}

1743: 

1744: We now turn to our numerical analysis of the series for punctured polygons.

1745: In section~\ref{sec:stair_min} we use our series to determine numerically the 

1746: {\em exact} area moment generating functions for minimally punctured

1747: staircase polygons with up to 5 punctures and $k\leq 10$. The resulting

1748: exact expressions for the leading amplitudes are in complete agreement with

1749: the formula derived in Proposition~\ref{prop:minpunc}. In section \ref{sec:stair_fix} we

1750: extend the study to staircase polygons with one puncture of fixed size and two punctures 

1751: of fixed combined size. Again we find the exact generating functions and 

1752: confirm the formula for the amplitude given in Theorem~\ref{theo:bounded}.

1753: Next, in section~\ref{sec:stair_one}, we analyse area moments for staircase polygons with 

1754: a single puncture of arbitrary size. Guided by results obtained from an analysis of the

1755: conjectured exact ODE \cite{GJ06b} satisfied by the perimeter generating function

1756: we carry out a careful numerical analysis of the area moment coefficients.

1757: This allows to obtain accurate estimates for the leading amplitudes

1758: and we confirm the results of Theorem~\ref{theo:arb} to at least 15 significant digits.

1759: Then, in section~\ref{sec:stair_two} we extend our study to staircase polygons

1760: with two staircase punctures of arbitrary size and we again find good agreement 

1761: with the exact results. Intriguingly, we find in all of the above cases, that

1762: the amplitude of the leading correction term is a constant times the

1763: corresponding amplitude with one less puncture. Finally, in section~\ref{sec:sapana}

1764: we present the results of our analysis of self-avoiding polygons  with one and two punctures 

1765: (minimal as well as arbitrary). In this case the numerical evidence is not 

1766: quite as convincing, but we do find that the numerical estimates agree with the exact formulas

1767: to at least 3--4 significant digits.

1768: 

1769: 

1770: \subsection{Staircase polygons with minimal punctures \label{sec:stair_min}}

1771: 

1772: In \cite{GJWE00} it was found that the half-perimeter generating function of staircase

1773: polygons with a single minimal puncture is: 

1774: 

1775: \begin{equation}\label{eq:St1p}

1776: \StM{1}(x) = \frac{1-8x+20x^2-16x^3+2x^4}{2(1-4x)}- \frac{1-6x+10x^2-4x^3}{2\sqrt{1-4x}}.

1777: \end{equation}

1778: This result is also derivable from Eq.~(\ref{sp1}), which gives a functional equation 

1779: for the area-perimeter generating function. 

1780: It is thus plausible to expect that the generating function of staircase polygons

1781: with $r$ minimal punctures is of a similar form

1782: 

1783: 

1784: \begin{equation}\label{eq:Stmin}

1785: \StM{r}(x) = \frac{A_r(x)+B_r(x)\sqrt{1-4x}}{(1-4x)^{\gamma_r}},

1786: \end{equation}

1787: where $A_r(x)$ and $B_r(x)$ are polynomials and $\gamma_r = (3r-1)/2$. We find this to be correct 

1788: for all the cases we have enumerated that is up to $r=5$:

1789: 

1790: 

1791: \begin{eqnarray*}

1792: 2A_2(x) &=& {x}^{2}-26{x}^{3}+228{x}^{4}-906{x}^{5}+1709{x}^{6}-1378{x}^{7}+322{x}^{8},\\

1793: 2B_2(x) &=& -{x}^{2}+24\,{x}^{3}-182\,{x}^{4}+586\,{x}^{5}-815\,{x}^{6}+404\,{x}^{7}-32\,{x}^{8}.\\

1794: A_3(x) &=& {x}^{2}-22\,{x}^{3}+197\,{x}^{4}-924\,{x}^{5}+2545\,{x}^{6}-5374\,{x}^{7}+13828\,{x}^{8} \\

1795: &&\!-33634\,{x}^{9}\!+46027\,{x}^{10}\!-24746\,{x}^{11}\!+612\,{x}^{12}\!+256\,{x}^{13}\!+256\,{x}^{14},\\

1796: B_3(x) &=& -{x}^{2}+20\,{x}^{3}-159\,{x}^{4}+642\,{x}^{5}-1509\,{x}^{6}+3176\,{x}^{7}-9040\,{x}^{8}\\

1797: &&+19254\,{x}^{9}-18943\,{x}^{10}+4968\,{x}^{11}+768\,{x}^{12}+256\,{x}^{13}.\\

1798: 2A_4(x) &=& 2\,{x}^{2}-60\,{x}^{3}+809\,{x}^{4}-6564\,{x}^{5}+36321\,{x}^{6}-146436\,{x}^{7}\\

1799: &&+439283\,{x}^{8}-960070\,{x}^{9}+1485167\,{x}^{10}-1823356\,{x}^{11}\\

1800: &&+2728708\,{x}^{12}-4441406\,{x}^{13}+4054296\,{x}^{14}-932228\,{x}^{15}\\

1801: &&-298318\,{x}^{16}-143360\,{x}^{17}+16384\,{x}^{18}-32768\,{x}^{19},\\

1802: 2B_4(x) &=& -2\,{x}^{2}+56\,{x}^{3}-701\,{x}^{4}+5266\,{x}^{5}-26987\,{x}^{6}+100694\,{x}^{7}\\

1803: &&-276415\,{x}^{8}+537888\,{x}^{9}-727683\,{x}^{10}+889018\,{x}^{11}\\

1804: &&-1536634\,{x}^{12}+2199158\,{x}^{13}-1289388\,{x}^{14}-47472\,{x}^{15}\\

1805: &&+26880\,{x}^{16}+50176\,{x}^{17}-6144\,{x}^{18}+8192\,{x}^{19}.\\

1806: 2A_5(x) &=& 2\,{x}^{2}-76\,{x}^{3}+1343\,{x}^{4}-14776\,{x}^{5}+114384\,{x}^{6}-666240\,{x}^{7}\\

1807: &&+3036602\,{x}^{8}-11071408\,{x}^{9}+32642310\,{x}^{10}-77911156\,{x}^{11}\\

1808: &&+148630330\,{x}^{12}-220310536\,{x}^{13}+250700412\,{x}^{14}\\

1809: &&-250317844\,{x}^{15}+290657417\,{x}^{16}-309183568\,{x}^{17}\\

1810: &&+150313538\,{x}^{18}+21743832\,{x}^{19}-15222464\,{x}^{20}+449152\,{x}^{21}\\

1811: &&-3828224\,{x}^{22}-2844672\,{x}^{23}+974848\,{x}^{24}-819200\,{x}^{25},\\

1812: 2B_5(x) &=& -2\,{x}^{2}+72\,{x}^{3}-1203\,{x}^{4}+12506\,{x}^{5}-91510\,{x}^{6}+504084\,{x}^{7}\\

1813: &&-2171612\,{x}^{8}+7467208\,{x}^{9}-20683474\,{x}^{10}+46059704\,{x}^{11}\\

1814: &&-80841764\,{x}^{12}+107986392\,{x}^{13}-111525400\,{x}^{14}\\

1815: &&+114888220\,{x}^{15}-142562573\,{x}^{16}+122527230\,{x}^{17}\\

1816: &&-24478856\,{x}^{18}-17117496\,{x}^{19}-533632\,{x}^{20}-2988544\,{x}^{21}\\

1817: &&-808960\,{x}^{22}+401408\,{x}^{23}-819200\,{x}^{24}.

1818: \end{eqnarray*}

1819: 

1820: Likewise the generating functions for

1821: the $k$'th area moment, $\StMM{r}{k}(x)$, is also of the form (\ref{eq:Stmin})

1822: \begin{equation}\label{eq:Stminmom}

1823: \StMM{r}{k}(x) = \frac{A_{r,k}(x)+B_{r,k}(x)\sqrt{1-4x}}{(1-4x)^{\gamma_{r+k}}}.

1824: \end{equation}

1825: We find that the degrees of the polynomials $A_{r,k}(x)$ and $B_{r,k}(x)$ are less than $5r+2k$ 

1826: for $r\leq5$ and $k\leq 10$. In particular we have:

1827: 

1828: 

1829: \begin{eqnarray}\label{eq:St1p1m}

1830: \StMM{1}{1}(x) =& \frac{1-14x+72x^2-162x^3+145x^4-34x^5+2x^6}{(1-4x)^{5/2}} \nonumber \\

1831: &-\frac{1-12x+50x^2-82x^3+43x^4-4x^5}{(1-4x)^2}.

1832: \end{eqnarray}

1833: 

1834: From these solutions we then calculate the exact leading amplitudes 

1835: and indeed we find that

1836: \begin{equation}

1837: \La{r}{k}=A_{k,r}(x_c)/\Gamma (\gamma_{r+k})

1838: =\frac{(-1)^{k+r}(k+r)! x_c^{2r}f_{k+r}}{r! x_c^{\gamma_{k+r}}\Gamma(\gamma_{k+r})}.

1839: \end{equation}

1840: in complete agreement with Eq.~(\ref{eq:relation}).

1841: 

1842: We have also looked at the amplitudes $\Lb{r}{k}=B_{k,r}(x_c)/\Gamma (\gamma_{r+k}-\frac12)$ 

1843: of the correction terms and find, quite remarkably, that they are given simply in terms

1844: of the leading amplitudes with one less puncture:

1845: 

1846: \begin{equation}\label{eq:StMB}

1847: \Lb{r}{k}=-\frac18 \La{r-1}{k}.

1848: \end{equation}

1849: 

1850: 

1851: \subsection{Staircase polygons with staircase punctures of fixed size \label{sec:stair_fix}}

1852: 

1853: Next we examine the more general case of staircase polygons with punctures of fixed size. In the

1854: case of one puncture we thus look at staircase polygons with a staircase hole of half-perimeter 

1855: $s$ while in the case of two punctures we look at staircase polygons with two staircase holes 

1856: whose half-perimeters sum to $s$. As in the previous section we expect the generating

1857: functions to be of the form

1858: \begin{equation}\label{eq:Stfix}

1859: \PPM{r,s}{k}(x) = \frac{A_{r,s,k}(x)+B_{r,s,k}(x)\sqrt{1-4x}}{(1-4x)^{\gamma_{r+k}}}.

1860: \end{equation}

1861: We find this to be true with the degree of the polynomials less than $5r+2(k+s)$.

1862: 

1863: For once punctured polygons we calculated the generating functions for $s \leq 25$ and $k \leq 10$.

1864: In Theorem~\ref{theo:bounded} we proved that the leading amplitude 

1865: $\La{1,s}{k}= A_{1,s,k}(x_c)/\Gamma(\gamma_{k+1})= A_{k+1}x_c^s p_s$,

1866: where $p_s$ is the number of staircase polygons of half-perimeter $s$. This formula is naturally

1867: confirmed by our numerical results. Of more interest is the sub-leading amplitude $\Lb{1,s}{k}$.

1868: We find that the result for minimally punctured polygons generalises to this case and

1869: $\Lb{1,s}{k} = -b_{1,s} A_k$. We also find that the integer sequence 

1870: $d_s=8^{s-1}b_{1,s}=1,5,29,182,\ldots$ is given by the recurrence:

1871: $$8s^2d_s+(s+3)(7s+10)d_{s+1}-(s+3)(s+2)d_{s+2}=0, d_1=0, d_2=1.$$

1872: We note that $d_s$ grows like $8^s$ so that $b_{1,s}$ grows no faster than a polynomial in $s$.

1873: 

1874: For twice punctured polygons we calculated the generating functions for $s \leq 10$ and $k \leq 10$.

1875: As a consequence of theorem~\ref{theo:bounded} the leading amplitude is given by

1876: $\La{2,s}{k}= A_{2,s,k}(x_c)/\Gamma(\gamma_{k+2})= A_{k+2}x_c^s \sum_{t=2}^{s-2}p_{s-t}p_t$.

1877: And again we find that $\Lb{2,s}{k} = -b_{2,s} A_{k+1}$, though in this case we haven't

1878: found a recurrence for the integer sequence $2\times 8^{s-2}b_{2,s}=1,9,69,510, \ldots$.

1879: 

1880: 

1881: \subsection{Staircase polygons with a single staircase puncture \label{sec:stair_one}}

1882: 

1883: We now turn to the analysis of the  area moments of staircase polygons with a single staircase 

1884: puncture of arbitrary size (1-punctured staircase polygons for short).

1885: In a recent paper \cite{GJ06b} we reported on work which led to an exact Fuchsian linear

1886: differential equation of order 8 apparently satisfied by the half-perimeter generating function, 

1887: $\PP{1}(x) = \sum_{m\geq 0} p_m^{(1)} x^m$, for 1-punctured staircase polygons 

1888: (that is $\PP{1} (x)$ is one of the solutions of the ODE, expanded around the origin). 

1889: Our analysis of the ODE showed that the dominant singular behaviour is

1890: \begin{equation}\label{eq:S1}

1891: \PP{1}(x) \sim \frac{A(x)}{(1-4x)}  + \frac{B(x) + C(x) \log(1-4x)}{\sqrt{1-4x}}+D(x) (1+4x)^{13/2}.

1892: \end{equation}

1893: The functions $A(x)$--$D(x)$ are regular in the disc $|x| \le 1/4$.

1894: In addition there were a pair of singularities on the imaginary axis at $x=\pm i/2$, 

1895: and at the roots of $1+x+7x^2$. Note that the absolute value of these singularities exceeds

1896: $1/4$ and so their contributions to the asymptotic growth of the series coefficients are 

1897: exponentially suppressed. 

1898: 

1899: 

1900: We expect that the area moment generating functions, $\PPM{1}{k}(x)$, should have a similar 

1901: critical behaviour to that of (\ref{eq:S1}). Indeed our analysis using 

1902: differential approximants \cite{AJG89a} revealed that at $x=x_c=1/4$ there

1903: is a triple root with exponents $-\gamma_{k+1}$ and  $-\gamma_{k+1}\!+\!1/2$ (twice)

1904: which is indicative that the behaviour is

1905: \begin{equation}\label{eq:S1mom}

1906: \PPM{1}{k}(x) \sim \frac{A(x)+[B(x) + C(x) \log(1-4x)]\sqrt{1-4x}}{(1-4x)^{3k/2+1}}.

1907: \end{equation}

1908: However, the behaviour at the singularity $x=x_-=-1/4$ is a little more complicated.

1909: For the first area moment we find that there is a double root with exponents $5$ and

1910: $13/2$, while for the second area moment we find a triple root at $x_-$ with

1911: exponents $7/2$, $5$ and $13/2$. For higher moments the behaviour is consistent 

1912: with a triple root with exponents $(13-3k)/2$, $(10-3k)/2$ and $(7-3k)/2$. That is, the

1913: value of the leading exponent decreases by $3k/2$ and there is in addition a non-analytic 

1914: correction to scaling with exponent $3/2$. The upshot of this analysis is that

1915: the  asymptotic behaviour of the coefficients of $\PPM{1}{k}(x)$ should be given by

1916: \begin{equation}\label{eq:S1co}

1917:  [x^m]\PPM{1}{k}(x)\! \sim\! 4^m\! \left ( \! \sum_{j=0} \!\!

1918:  m^{3k/2-j}\! \left  (\! a_j\!+\!\frac{1}{\sqrt{m}}\left [ b_j \!+\!c_j \log (m) \right ]\! \right ) 

1919: \!\!+\!(-1)^m \! \sum_{j=0}\! d_j m^{(-15+3k-j)/2}\! \right )\!,

1920: \end{equation}

1921: where we have ignored the contributions from singularities in the complex plane with absolute values

1922: exceeding $1/4$. Estimates for the amplitudes were obtained by fitting the coefficients

1923: $p_m^{(1,k)}=[x^m]\PPM{1}{k}(x)$ to the form given above using an increasing number of 

1924: amplitudes. `Experimentally' we find we need about the same total number of terms at 

1925: $x_c$ and $x_-$. So in the fits we used the terms with amplitudes $a_i$, $b_i$ and $c_i$, 

1926: $i=0,\ldots,K$ and $d_i$, $i=0,\ldots,3K$. Going only to $K$ with the $d_i$ amplitudes 

1927: results in much poorer convergence and going beyond $3K$ leads to no improvement. So we 

1928: use the $6K+4$ terms $p_m^{(1,k)}$ with $m$ ranging from $M$ to $M-6K-3$ and solve the

1929: resulting  system of $6K+4$ linear equations. 

1930: 

1931: \begin{figure}[htbp]

1932:   \centering

1933:   \includegraphics[scale=0.95]{St1_ampl}

1934:   \caption{\label{fig:1StA} The relative precision of the estimates for the leading amplitude,

1935: $|a_0 - \La{1}{k}|/\La{1}{k}$, against $1/M$ for the first (top left panel), second

1936: (top right), fifth (bottom left) and tenth (bottom right) area moments.}

1937: \end{figure}

1938: 

1939: We compare the amplitude estimates to the predictions in Section~\ref{sec:scaling_arbitrary}

1940: and we find that the estimated leading amplitude $a_0$ is given by the exact formula

1941: \begin{equation}\label{eq:StA}

1942: \La{r}{k}=\frac{(-1)^{k+r}(k+r)! x_c^r f_{k+r}}{r! x_c^{\gamma_{k+r}}\Gamma(\gamma_{k+r})},

1943: \end{equation}

1944: which agrees with Eq.~(\ref{eq:Aka}) since the critical half-perimeter generating,

1945: see Eq.~(\ref{eq:StGf}), for staircase polygons is $\PGf(x_c)=1/4=x_c$.

1946: In \cite{GJ06b} we studied the normalised coefficients

1947: $r_m = p_{m+8}^{(1)}/4^m.$ Using the recurrence relations for $p_m^{(1)}$ (derived from the ODE) 

1948: it is easy and fast to generate many more terms $r_m$. We generated the first 100000 terms and saved 

1949: them as floats with 500 digit accuracy. We found (to better than 100 digits) that the leading

1950: amplitude of the normalised series was $\tilde{a}_0=1024$. Going back to the normalisation 

1951: used in this paper we find $a_0=1024/4^8=1/64$ in agreement with formula~(\ref{eq:StA}).

1952: In figure~\ref{fig:1StA} we plot the relative precision of the estimates for the amplitude,

1953: $|a_0 - \La{r}{k}|/\La{r}{k}$ against $1/M$ for some of the area moments.

1954: Recall that for $k=1$ we have a series of 352 non-zero terms, for $k=2$ we have 292 terms

1955: and for $k=3$--10 we have 246 terms. 

1956: In all cases (including for area moments not shown) the relative precision of the estimate 

1957: $a_0$ is better than $10^{-15}$ for $K=10$. For the first area moment, where we have a longer

1958: series, the precision is even more impressive, being better than $10^{-20}$. So in all cases we

1959: can confirm the exact prediction (\ref{eq:StA}) for the amplitude to at least 15 significant

1960: digits.

1961: 

1962: 

1963: \begin{figure}[htbp]

1964:   \centering

1965:   \includegraphics[scale=0.95]{St1_logampl}

1966:   \caption{\label{fig:1StC} The relative precision of the estimates for the amplitude,

1967: $|c_0 - \Lc{1}{k}|/\Lc{1}{k}$, against $1/M$ for the zeroth (leftmost panel) 

1968: and tenth (rightmost panel) area moments.}

1969: \end{figure}

1970: 

1971: In the previous section we saw  that for minimally punctured staircase polygons

1972: the amplitude of the correction term is just a constant times

1973: the leading amplitudes with one less puncture, see Eq.~(\ref{eq:StMB}). It is thus

1974: natural to ask if something similar happens in the more general case.

1975: And indeed we find numerically that the amplitude $\Lc{1}{k}$ of the dominant 

1976: correction in (\ref{eq:S1mom}) (the one proportional to the log-term) giving 

1977: rise to the $c_j$-terms in (\ref{eq:S1co}) are

1978: \begin{equation}\label{eq:S1C}

1979: \Lc{1}{k} = - \frac{3\sqrt{3}}{8\pi} A_k.

1980: \end{equation}

1981: In figure~\ref{fig:1StC} we have plotted the relative error between the estimate

1982: of $c_0$ and the predicted value $\Lc{1}{k}$ for $k=0$ and 10. The estimates for the

1983: other moments are very similar with the accuracy of the agreement diminishing with

1984: higher moments. So in all cases (\ref{eq:S1C}) has been confirmed to better than 10 

1985: digit accuracy. 

1986: 

1987: For the sub-dominant correction term the amplitude $\Lb{1}{k}$, as approximated

1988: by the term $b_0$ in (\ref{eq:S1co}), is not simply related to $A_k$ and in fact

1989: it even changes sign as $k$ is increased.

1990: 

1991: 

1992: \subsection{Staircase polygons with two staircase punctures \label{sec:stair_two}}

1993: 

1994: In this section we report on our analysis of the series for staircase polygons with

1995: two punctures of arbitrary size. Our first task is to work out the singularity

1996: structure of the perimeter generating function $\PP{2}(x)$ (for which we have

1997: a series with 240 non-zero terms). From the general

1998: considerations of Section~\ref{sec:punc} we expect a singularity at $x=x_c=1/4$ with

1999: exponent $-5/2$, but given the quite complicated singularity structure of  $\PP{1}(x)$,

2000: as detailed in Eq.~(\ref{eq:S1}) and below, we would expect similar complications 

2001: for  $\PP{2}(x)$. We analysed  the series for  $\PP{2}(x)$ using differential

2002: approximants \cite{AJG89a}. This analysis revealed that there is a triple root at $x=x_c=1/4$

2003: and the exponents had values $-2.499(1)$, $-2.070(5)$ and $-1.78(1)$.

2004: So despite having a series of 240 terms it is still very difficult

2005: to pin down the exponents accurately. Given the values quoted above two possible

2006: scenarios present themselves. Either the exponents have the exact values $-5/2$, $-2$ and 

2007: $-2$ or they have the exact values $-5/2$, $-2$ and $-7/4$. The behaviour of

2008: $\PP{1}(x)$ would support the first of these scenarios and below we shall present

2009: evidence from the analysis of the asymptotic form of the coefficients which strongly

2010: supports this behaviour. We also find a double root singularity at $x=x_-=-1/4$

2011: with exponent estimates consistent with the exact values 5 and 11/2. 

2012: In addition there are several conjugate pairs of singularities

2013: in the complex plane. The most important of these are at $x=(-1 \pm i \sqrt{3})/8$,

2014: which has magnitude $1/4$ and thus lies equidistant from the origin to $x_c$. So unlike

2015: the situation for once punctured staircase polygons we cannot ignore the complex singularities.

2016: The exponent estimate at this singularity is consistent with the exact value $33/2$. 

2017: The singularities at $x=(-1 \pm i \sqrt{3})/8$ 

2018: are the roots of the polynomial $1+4x+16x^2$ which would indicate that the generating function 

2019: contains a term $\sim E(x)(1+4x+16x^2)^{33/2}$. Finally, we find some singularities

2020: with magnitude greater than $1/4$. There are singularities at $x=(1 \pm i \sqrt{3})/6$

2021: (which has magnitude $1/3$) with an exponent $33/2$ (we note that these are  the roots

2022: of ($1-3x+9x^2$)) and at $x=\pm i/2$ (magnitude $1/2$) with

2023: an exponent consistent with the value $5$. We also find weak evidence that $\PP{2}(x)$, 

2024: just as $\PP{1}(x),$ has a pair of singularities at the roots of $1+x+7x^2$.

2025: 

2026: As noted above we need to include the contribution from a conjugate pair of complex 

2027: singularities to the asymptotic form of the coefficients. In general this is not as 

2028: straight-forward a task as dealing with singularities on the real axis. In \cite{IJ06a} 

2029: we examined the generating function of self-avoiding walks on the honeycomb lattice 

2030: which has a pair of singularities on the imaginary axis at $x=\pm i/\tau$, arising 

2031: from a term of the form $H(x)(1+\tau^2 x^2)^{-\eta}$. This typically produces coefficients 

2032: which change sign according to a $++--$ pattern. This can be accommodated by including 

2033: terms of the form

2034: 

2035: $$

2036: \sim \tau^m m^{\eta-1} \sum_{j\geq 0} (-1)^{\lfloor (m+j)/2 \rfloor} h_j/m^j

2037: $$

2038: in fitting to the asymptotic form of the coefficients.

2039: 

2040: Note that the analysis in \cite{IJ06a} clearly demonstrated that, as is done above, 

2041: one has to shift the sign-pattern by $j$ when terms proportional to $1/m^j$ are 

2042: considered. Terms arising from other complex conjugate pairs of singularities can 

2043: give rise to much more complicated sign-patterns. In order to handle such cases we 

2044: simply form the Taylor expansion of the simplest possible term arising from the 

2045: singularity and take the sign of the appropriate coefficient. Specifically in 

2046: order to include terms proportional to $1/m^j$, when looking at the coefficients 

2047: $[x^m]\PP{2}(x)$, we take the sign to be the sign of the coefficient of $x^{m+j}$ 

2048: in the Taylor expansion of the function $(1+4x+16x^2)^{33/2}$. We use the notation 

2049: $\Sign{(1\!+\!4x\!+\!16x^2)^{33/2}}{m\!+\!j}$ for this operation.

2050: 

2051: 

2052: The singularity structure revealed above leads us to fit the coefficients 

2053: to a form, which is appropriate if the first scenario (exponents at $x=x_c=1/4$ 

2054: are $-5/2$, $-2$ and $-2$) is correct 

2055: \begin{eqnarray}\label{eq:St2Log}

2056:  [x^m]\PP{2}(x)\! =&\! 4^m\! \left (  \sum_{j=0}^K \!

2057:  m^{3/2-j}\! \left  [ a_j\!+\!\frac{1}{\sqrt{m}}\left [ b_j \!+\!c_j \log (m) \right ]\! \right ] 

2058: +(-1)^m \sum_{j=0}^{3K}\! d_j m^{-6-j/2} \right .\nonumber \\

2059: &\left .+ \sum_{j=0}^{3K}\! e_j \Sign{(1\!+\!4x\!+\!16x^2)^{33/2}}{m\!+\!j} m^{-35/2+j}\! \right ),

2060: \end{eqnarray}

2061: where we have ignored the singularities with magnitude exceeding $1/4$. 

2062: We also examine the alternative form appropriate if the second scenario 

2063: (exponents at $x=x_c=1/4$ are $-5/2$, $-2$ and $-7/4$) is correct 

2064: \begin{eqnarray}\label{eq:St2Exp}

2065:  [x^m]\PP{2}(x)\! =&\! 4^m\! \left ( \sum_{j=0}^K \!

2066:  m^{3/2-j}\left [ a_j\!+\!\frac{b_j}{m^{1/2}}\!+\!\frac{c_j}{m^{3/4}}\! \right ] 

2067:  +(-1)^m \sum_{j=0}^{3K}\! d_j m^{-6-j/2}\!\right . \nonumber \\

2068: &\left . + \sum_{j=0}^{3K}\! e_j \Sign{(1\!+\!4x\!+\!16x^2)^{33/2}}{m\!+\!j} m^{-35/2+j}\! \right ). 

2069: \end{eqnarray}

2070: 

2071: \begin{figure}[htbp]

2072:   \centering

2073:   \includegraphics[scale=0.95]{St2_ampl}

2074:   \caption{\label{fig:2StA} Estimates of the leading amplitude $a_0$ and the amplitude $c_0$ of the

2075:   sub-dominant term.

2076:   The top panels show the results from fitting to the form~(\protect{\ref{eq:St2Log}})

2077:   while the bottom panels are results from fitting to the form~(\protect{\ref{eq:St2Exp}}).

2078:   The straight line is the exact value of this amplitude $\La{2}{0}$}

2079: \end{figure} 

2080: 

2081: In figure~\ref{fig:2StA} we have plotted the resulting estimates for the leading amplitude

2082: $a_0$, which we expect are given by the exact value  $\La{2}{0} =\frac{5}{3072\sqrt{\pi}}$, and the 

2083: amplitude $c_0$ of the sub-dominant terms from the two alternative

2084: asymptotic forms. First we focus on the sub-dominant terms. In the top right

2085: panel we show the estimates when fitting to the form~(\ref{eq:St2Log}) where we have a

2086: sub-dominant term proportional to $m \log (m)$ while the bottom right panel shows the

2087: estimates obtained when fitting to the form~(\ref{eq:St2Exp}) where the sub-dominant

2088: term is $m^{3/4}$ (in both cases the dominant term is proportional to $m^{3/2}$ with

2089: a second sub-dominant term proportional to $m$). In the bottom right panel we note that

2090: the estimates for the amplitude $c_0$ of the term $m^{3/4}$ seems to diverge. As $K$

2091: is increased rather than settle down we find that the slope of the curves of the estimates

2092: plotted vs. $1/M$ {\em increases} which is the opposite of what we would expect if we

2093: are fitting to the correct asymptotic form. We take this as firm numerical evidence that 

2094: (\ref{eq:St2Exp}) is incorrect. This could also explain why the corresponding estimates of 

2095: $a_0$ (bottom left panel) don't appear to converge to the predicted exact value. In

2096: contrast the estimates for $c_0$ of the term $m\log (m)$ (top right panel) from the

2097: form~(\ref{eq:St2Log}) do seem to converge and the slopes of the estimates plotted vs. $1/M$ 

2098: decrease with $K$. In this case the estimates for $a_0$ (top left panel) clearly can be

2099: extrapolated to a value consistent with the predicted exact value. Note further

2100: that the top left panel has a resolution along the ordinate which is a factor 5 higher than

2101: in lower right panel so the estimates when fitting to the form~(\ref{eq:St2Log}) are

2102: much more tightly converged. The conclusion is that the numerical evidence clearly

2103: favours the asymptotic form~(\ref{eq:St2Log}) and we believe this to be (if not entirely

2104: correct at least a very good approximation to) the true asymptotic form of the

2105: coefficients of the generating function $\PP{2}(x)$ for twice punctured staircase polygons.

2106: 

2107: 

2108: \begin{figure}[htbp]

2109:   \centering

2110:   \includegraphics[scale=0.95]{St2_mom}

2111:   \caption{\label{fig:2StM} Estimates of the leading amplitude 

2112:   $\La{2}{k}$ of various area moments

2113:   obtained by fitting to the form~(\protect{\ref{eq:St2Mom}}).

2114:   The straight line is the exact value of this amplitude.}

2115: \end{figure} 

2116: 

2117: Now that we have settled the question of the correct singularity structure of $\PP{2}(x)$

2118: we turn our attention to the analysis of the area moment generating function $\PPM{2}{k}(x)$.

2119: As for once punctured staircase polygons we find that the leading exponent at all singularities

2120: decreases by $3k/2$. So in order to estimate the leading amplitude $a_0$ we fit to the form

2121: \begin{eqnarray}\label{eq:St2Mom}

2122:  [x^m]\PPM{2}{k}(x)\! =&\! 4^m\! \left (  \sum_{j=0}^K \!

2123:  m^{3(1+k)/2-j}\! \left  [ a_j\!+\!\frac{1}{\sqrt{m}}\left [ b_j \!+\!c_j \log (m) \right ]\! \right ] 

2124: +(-1)^m \sum_{j=0}^{3K}\! d_j m^{(-12+3k+j)/2} \right .\nonumber \\

2125: &\left .+ \sum_{j=0}^{3K}\! e_j \Sign{(1\!+\!4x\!+\!16x^2)^{33/2}}{m\!+\!j} m^{-(35-3k+j)/2}\! \right ),

2126: \end{eqnarray}

2127: where for simplicity our notation suppresses the $k$-dependence of the amplitudes.

2128: Recall that for $k=1$ and 2 we have series with 214 terms and for $k=3$--10 we

2129: have 140 terms.

2130: We compare the amplitude estimates to the predictions of Section~\ref{sec:scaling}.

2131: The leading amplitude $a_0=\La{2}{k}$ is given by the exact 

2132: formula~(\ref{eq:StA}). In figure~\ref{fig:2StM} we show the estimates for the 

2133: leading amplitudes for area moments with $k=1$, 2, 5, and 10.

2134: In all cases the amplitudes estimates appears to converge to the predicted exact

2135: value and agreement is found to at least 3 significant digits.

2136: 

2137: 

2138: \begin{figure}[htbp]

2139:   \centering

2140:   \includegraphics[scale=0.95]{St2_logampl}

2141:   \caption{\label{fig:2StC} The ratio $c_0/\La{1}{k}$ for the zeroth area moment 

2142: (leftmost panel) and several different area moments (rightmost panel).}

2143: \end{figure}

2144: 

2145: Finally we turn our attention to the amplitude $\Lc{2}{k}$ of the dominant correction term.

2146: In the leftmost panel of figure~\ref{fig:2StC} we have plotted the ratio between 

2147: the estimated amplitude $c_0$ and the predicted value of $\La{1}{k}$ for $k=0$ using

2148: several cut-offs. In the rightmost panel we show the same ratio but for

2149: several different moments using the cut-off $K=6$. These plots are consistent

2150: with $\Lc{2}{k} \propto \La{1}{k}$ with the constant of proportionality being

2151: $-0.212(2)$.

2152: 

2153: \subsection{Punctured self-avoiding polygons \label{sec:sapana}}

2154: 

2155: \begin{table}

2156: \begin{center}

2157: \small

2158: \begin{tabular}{llll}

2159: \hline\hline

2160: \multicolumn{1}{c}{Degrees} & \multicolumn{3}{c}{Once punctured $k=0$} \\

2161: \hline

2162: $[8,9,9,10]$ &  $ -0.002022  -0.023965i$ & $ -0.002022+  0.023965i$ & $  0.518101$ \\ 

2163:  $[8,9,9,11]$ &  $ -0.003847+  0.034768i$ &  $ -0.003847  -0.034768i$  & $  0.528973$ \\ 

2164: $[8,9,9,12]$ &  $ -0.004553  -0.038104i$ & $ -0.004553+  0.038104i$ & $  0.533388$ \\ 

2165:  $[8,9,10,10]$ &  $ -0.001830  -0.023143i$ & $ -0.001830+  0.023143i$ & $  0.515916$ \\ 

2166:  $[8,9,10,11]$ &  $ -0.044131$ &  $  0.060194$ &  $  0.459558$ \\ 

2167:  $[8,9,11,10]$ &  $ -0.006365  -0.045090i$ & $ -0.006365+  0.045090i$ & $  0.547479$ \\ 

2168:  $[8,10,9,10]$ &  $ -0.007054  -0.047578i$ & $ -0.007054+  0.047578i$ & $  0.552445$ \\ 

2169:  $[8,10,9,11]$ &  $ -0.009398  -0.055135i$ & $ -0.009398+  0.055135i$ & $  0.570859$ \\ 

2170:  $[8,10,10,10]$ &  $ -0.012495  -0.063935i$ & $ -0.012495+  0.063935i$ & $  0.595957$ \\ 

2171: \hline\hline

2172: \end{tabular}

2173: \end{center}

2174: \caption{\label{tab:SAP1pM0} 

2175: Biased estimates of the critical exponents of once punctured SAPs.}

2176: \end{table}

2177: 

2178: Before proceeding to the estimation of the amplitudes we briefly have a look

2179: at the critical behaviour of the area moment generating functions for

2180: punctured self-avoiding polygons. In \cite{GJWE00} we analysed the behaviour of

2181: $\PPM{r}{k}(x)$  and found  the critical behaviour to be consistent with

2182: $\PPM{r}{k}(x) \sim A(x)+B(x)(x-x_c)^{-\gamma_{k+r}}+ C(x)(x-x_c)^{-\gamma_{k+r}+1/2}$,

2183: where $\gamma_{j} = 3j/2-3/2$.

2184: From previous work \cite{JG99,IJ03} we have very precise estimates for $x_c,$ which

2185: is indistinguishable from the positive root of the polynomial $581x^2+7x-13$, that is,

2186: $x_c=0.1436806292\ldots$. Using this value for $x_c$

2187: we form a $K^{th}$-order {\em biased} differential approximant (DA) to $\PPM{r}{k}(x)$ 

2188: by matching the coefficients in the polynomials $Q_i(x)$ of degree $N_i$

2189: so that (one) of the formal solutions to the homogeneous differential equation

2190: $$

2191: \sum_{i=0}^K (x-x_c)^iQ_i(x)(x\frac{{\rm d}}{{\rm d}x})^i \tilde{P}(x) = 0

2192: $$

2193: agrees with the first $M=\sum_i (N_i+1)$ series coefficients of $\PPM{r}{k}(x)$.

2194: We are thus `forcing' the DAs to have regular singular points at the origin and $x_c$.

2195: The critical exponents $\gamma_j$ ($j=1\ldots K$) are estimated from the indicial equation 

2196: at $x_c$ (note that due to the biasing, $x_c$ is root of order $K$). 

2197: If the true singular behaviour at $x_c$ implies a root of degree less than $K$ we

2198: expect that the `true' exponents will be quite well estimated and show

2199: little scatter while the `surplus' exponents will show a lot of random scatter.

2200: In the following we always use $K=3$ and denote the degrees of the polynomials

2201: $Q_i$ as $[N_3,N_2,N_1,N_0]$.

2202: 

2203: First we look at the perimeter generating function for once punctured SAPs (for

2204: which we have series with 42 terms).

2205: In this case we have $\gamma_1 =0$ and we thus expect a logarithmic singularity at $x_c$ 

2206: with a square-root correction term as argued in \cite{GJWE00}.

2207: In Table~\ref{tab:SAP1pM0} we list some exponent estimates obtained from the biased

2208: DAs. The exponent estimates are indeed consistent with the exact values $0,0,1/2$,

2209: which confirms the expected behaviour.

2210:   

2211: 

2212: \begin{table}

2213: \begin{center}

2214: \small

2215: \begin{tabular}{lllllll}

2216: \hline\hline

2217: \multicolumn{1}{c}{Degrees} & \multicolumn{3}{c}{Once punctured $k=2$} 

2218:                     & \multicolumn{3}{c}{Once punctured $k=5$} \\

2219: \hline

2220:  $[8,9,9,10]$ &  $ -3.000291$ &  $ -2.493921$ &  $ -1.154513$ 

2221:  &  $ -7.500838$ &  $ -6.976201$ &  $ -5.107975$ \\  

2222: $[8,9,9,11]$ &  $ -3.000158$ &  $ -2.496645$ &  $ -1.258606$  

2223:  &  $ -7.503030$ &  $ -6.927432$ &  $ -2.422194$ \\ 

2224:  $[8,9,9,12]$  &  $ -3.012422$ &  $ -2.421400$  &  $ -3.323742$

2225:   &  $ -7.500231$ &  $ -6.992184$ &  $ -5.461825$ \\ 

2226:  $[8,9,10,10]$ &  $ -2.999849$ &  $ -2.502858$ &  $ -1.447159$ 

2227:   & $ -7.517118$ &  $ -6.776728$  &  $ -8.010389$ \\ 

2228:  $[8,9,10,11]$ &  $ -3.000452$ &  $ -2.491160$ &  $ -1.022360$  

2229:   &  $ -7.500700$ &  $ -6.980331$ &  $ -5.251153$ \\ 

2230:  $[8,9,11,10]$ &  $ -3.000143$ &  $ -2.497148$ &  $ -1.307314$  

2231:   &  $ -7.500366$ &  $ -6.988804$ &  $ -5.406296$ \\ 

2232:  $[8,10,9,10]$ &  $ -2.999481$ &  $ -2.511762$ &  $ -1.668527$ 

2233:    &  $ -7.505629$ &  $ -6.879658$ &  $-14.20958$\\ 

2234:  $[8,10,9,11]$ &  $ -2.999906$ &  $ -2.501910$ &  $ -1.441981$  

2235:   &  $ -7.500535$ &  $ -6.984576$ &  $ -5.361394$ \\ 

2236:  $[8,10,10,10]$ &  $ -2.999647$ &  $ -2.507601$ &  $ -1.578352$  

2237:   &  $ -7.500946$ &  $ -6.974159$ &  $ -5.059683$ \\ 

2238: \hline\hline

2239: \end{tabular}

2240: \end{center}

2241: \caption{\label{tab:SAP1pM25} 

2242: Biased estimates of the critical exponents for the 2nd and 5th area moments of once punctured SAPs.}

2243: \end{table}

2244:  

2245: 

2246: \begin{table}

2247: \begin{center}

2248: \small

2249: \begin{tabular}{lllrlll}

2250: \hline\hline

2251: \multicolumn{1}{c}{Degrees} & \multicolumn{3}{c}{Twice punctured $k=0$} 

2252:                     & \multicolumn{3}{c}{Twice punctured $k=2$} \\

2253: \hline

2254:  $[7,8,8,9]$ &  $ -1.504250$ &  $ -0.968766$ &  $  6.594627$ 

2255:   &  $ -4.499111$ &  $ -4.006801$ &  $ -2.544930$ \\ 

2256:  $[7,8,8,10]$ &  $ -1.504232$ &  $ -0.968829$ &  $  5.029521$ 

2257:   &  $ -4.495871$ &  $ -4.039199$ &  $ -3.083192$ \\ 

2258:  $[7,8,8,11]$ &  $ -1.504209$ &  $ -0.968910$  &  $-60.90721$

2259:   &  $ -4.494258$ &  $ -4.048789$ &  $ -2.701322$ \\ 

2260:  $[7,8,9,9]$ &  $ -1.504210$ &  $ -0.968909$ &  $ 12.03204$ 

2261:   &  $ -4.494568$ &  $ -4.047486$ &  $ -2.836670$ \\ 

2262:  $[7,8,9,10]$ &  $ -1.504197$ &  $ -0.968951$ &  $ 32.28205$  

2263:   &  $ -4.493872$ &  $ -4.051402$ &  $ -2.672813$ \\ 

2264:  $[7,8,10,9]$ &  $ -1.504205$ &  $ -0.968923$ &  $ 22.27646$  

2265:   &  $ -4.508207$ &  $ -3.915977$ &  $ -1.950665$ \\ 

2266:  $[7,9,8,9]$ &  $ -1.504178$ &  $ -0.969017$ &  $ 18.18298$  

2267:   &  $ -4.495340$ &  $ -4.040712$ &  $ -2.812621$ \\ 

2268:  $[7,9,8,10]$ &  $ -1.504177$ &  $ -0.969020$ &  $ 18.81517$ 

2269:   &  $ -4.495136$ &  $ -4.041237$ &  $ -2.679530$ \\ 

2270:  $[7,9,9,9]$ &  $ -1.504176$ &  $ -0.969022$ &  $ 18.58757$ 

2271:   &  $ -4.498396$ &  $ -4.012876$ &  $ -2.698591$ \\ 

2272: \hline\hline

2273: \end{tabular}

2274: \end{center}

2275: \caption{\label{tab:SAP2pM02} 

2276: Biased estimates of the critical exponents for the 0th and 2nd area moments of twice punctured SAPs.}

2277: \end{table}

2278:  

2279: In Table~\ref{tab:SAP1pM25} we list some exponent estimates for the 2nd and 5th 

2280: area moments of once punctured SAPs. The exponents support the expectations for the

2281: leading and sub-dominant exponent. The third exponent is of no significance--the nature

2282: of the differential approximant forces a third exponent to have some value, but its lack

2283: of convergence suggests it is not, in fact present.  In Table~\ref{tab:SAP2pM02} we list exponent

2284: estimates for the 0th and 2nd area moments of twice punctured SAPs (for

2285: which we have series with 38 terms). Similar comments

2286: apply to the three columns of exponent estimates as were made for the 2nd area moments.

2287: In all cases we get a clear confirmation of the critical behaviour observed in \cite{GJWE00}.

2288: 

2289: Next we turn our attention to estimates for the critical amplitudes.

2290: Proposition~\ref{prop:minpunc} and Theorem~\ref{theo:arb} tells us that the critical amplitude

2291: of the $k$th area moment of self-avoiding polygons with $r$ (minimal or arbitrary)

2292: punctures is proportional to the critical amplitude of the  $(k+r)$th area moment 

2293: of unpunctured SAPs (the theorems also give the constants of proportionality). 

2294: In order to test these predictions numerically we analyse in this section data

2295: for SAPs with one and two punctures. In all cases we look at the ratio 

2296: $r_m=p_m^{(r,k)}/p_m^{(k+r)}$ which should approach the relevant constant of 

2297: proportionality. Given the critical behaviour outlined above we expect that these 

2298: amplitude ratios can be approximated quite well by the asymptotic form

2299: $$r_m = \sum_{j=0} a_j m^{j/2}$$

2300: So as in the analysis of punctured staircase polygons we obtain estimates for the 

2301: leading amplitude $a_0$, by fitting to the above form truncated at some level $K$, 

2302: and we then plot these estimates against $1/M$. 

2303: 

2304: 

2305: 

2306: \subsubsection{Minimal punctures}

2307: 

2308: According to Proposition~\ref{prop:minpunc} the amplitude of the $k$-th area moment of 

2309: SAPs with $r$ minimal punctures is 

2310: 

2311: \begin{equation}

2312: A_k^{(r)}=A_{k+r}x_c^{2r}/r!.

2313: \end{equation}

2314: 

2315: We first analyse the area moments of SAPs with a single minimal puncture. 

2316: In the left panel of figure~\ref{fig:SapM1A} we have plotted estimates for amplitude ratio

2317: $A_0^{(1)}/A_1$ with $7 \leq K \leq 10$. The prediction for this ratio is $x_c^2$, which

2318: is plotted as a straight line. The estimates obtained with $K=9$ and $10$ are indistinguishable

2319: from the predicted value at the resolution of the plot. Note that the `curvature' of

2320: the plotted values decreases as $K$ increases. We take this to be a very clear indication

2321: that the ratio $r_m$ is very well approximated by the assumed asymptotic form.

2322: In the right panel we plot the estimates for the amplitude ratio $A_k^{(1)}/A_{k+1}$ 

2323: for $0 \leq k \leq 9$ using the cut-off value $K=10$. For small values of $k$ these plots

2324: give firm numerical evidence for the correctness of  Proposition~\ref{prop:minpunc}. For higher

2325: values of $k$ (8 and 9 in particular) the evidence is not quite as firm though nothing

2326: in the plot would suggest a discrepancy with the predicted value. We again emphasise

2327: that in this case we have relatively short series of only 42 terms.

2328: 

2329: 

2330: 

2331: \begin{figure}[htbp]

2332:   \centering

2333:   \includegraphics[scale=1.0]{SapM1_amplpred}

2334:   \caption{\label{fig:SapM1A} Estimates of the amplitude ratio $A_k^{(1)}/A_{k+1}$ 

2335:    for self-avoiding polygons with one minimal puncture.}

2336: \end{figure}

2337: 

2338: 

2339: Next we analyse the area moments of SAPs with two minimal punctures. 

2340: The left panel of figure~\ref{fig:SapM2A} shows estimates for amplitude ratio

2341: $A_0^{(2)}/A_2$ with $7 \leq K \leq 10$. The prediction for the ratio, $x_c^4/2$,

2342: is plotted as a straight line.  Again we find that the estimates for $K=10$ are

2343: indistinguishable from the predicted value, though in this case the relative 

2344: resolution is coarser than in the previous plot. 

2345: In the right panel we plot estimates of the amplitude ratio $A_k^{(2)}/A_{k+2}$ 

2346: for $0 \leq k \leq 8$ using the cut-off value $K=10$. Again we find that our

2347: numerical analysis confirms the prediction to a high degree of confidence.

2348: Recall that in this case we have series with only 38 terms.

2349: 

2350: \begin{figure}[htbp]

2351:   \centering

2352:   \includegraphics[scale=1.0]{SapM2_amplpred}

2353:   \caption{\label{fig:SapM2A} Estimates of the amplitude ratio $A_k^{(2)}/A_{k+2}$ 

2354:   for self-avoiding polygons with two minimal punctures.}

2355: \end{figure}

2356: 

2357: 

2358: \subsubsection{Arbitrary punctures}

2359: 

2360: According to Theorem~\ref{theo:arb} the amplitude of the $k$-th area moment of 

2361: SAPs with $r$ arbitrary punctures is 

2362: 

2363: \begin{equation}

2364: A_k^{(r)}=A_{k+r}({\cal P}(x_c))^r/r!, 

2365: \end{equation}

2366: where $  {\cal P}(x_c)$ is the critical amplitude of the half-perimeter generating function.

2367: 

2368: The first step in our analysis is to obtain an accurate estimate of ${\cal P}(x_c)$.

2369: In \cite{JG99} we obtained the rather imprecise estimate ${\cal P}(x_c)\approx 0.036$ 

2370: by evaluating Pad\'e approximants to the generating function. Here we shall

2371: estimate ${\cal P}(x_c)$ directly from the perimeter data. We first tried the form

2372: %

2373: \begin{displaymath}

2374: \sum_{m=0}^M p_{m}x_c^m \sim  {\cal P}(x_c)+a_1/M^{1/2}+a_2/M+\cdots

2375: \end{displaymath}

2376: Using the first ten terms in this asymptotic expansion we found (with $M=55$) 

2377: that ${\cal P}(x_c)\approx 0.0362642$, but $a_1\approx 4.28\times10^{-9}$ and

2378: $a_2\approx -1.42\time 10^{-7}$, while $a_3 \approx -0.066$. We are therefore

2379: quite confident that $a_1=a_2=0$. Upon further analysis we found convincing evidence

2380: that the correct asymptotic form in fact is 

2381: 

2382: \begin{displaymath}

2383: \sum_{m=0}^M p_{m}x_c^m \sim  {\cal P}(x_c)+b_1/M^{3/2}+b_2/M^{5/2}+\cdots

2384: \end{displaymath}

2385: 

2386: 

2387: \begin{table}

2388: \begin{center}

2389: \begin{tabular}{cccc}

2390: \hline\hline

2391: $M$ & $K=8$ & $K=10$ & $K=12$ \\

2392: \hline

2393: 45  &   0.036264215181387

2394:     &   0.036264215181095

2395:     &   0.036264215180475 \\

2396: 46  &   0.036264215181343

2397:     &   0.036264215181088

2398:     &   0.036264215181354 \\

2399: 47  &   0.036264215181305

2400:     &   0.036264215181073

2401:     &   0.036264215180915 \\

2402: 48  &   0.036264215181271

2403:     &   0.036264215181060

2404:     &   0.036264215180994 \\

2405: 49  &   0.036264215181240

2406:     &   0.036264215181048

2407:     &   0.036264215180970 \\

2408: 50  &   0.036264215181214

2409:     &   0.036264215181038

2410:     &   0.036264215181001 \\

2411: 51  &   0.036264215181190

2412:     &   0.036264215181029

2413:     &   0.036264215180972 \\

2414: 52  &   0.036264215181168

2415:     &   0.036264215181021

2416:     &   0.036264215180972 \\

2417: 53  &   0.036264215181149

2418:     &   0.036264215181013

2419:     &   0.036264215180969 \\

2420: 54  &   0.036264215181131

2421:     &   0.036264215181006

2422:     &   0.036264215180967 \\

2423: 55  &   0.036264215181116

2424:     &   0.036264215181000

2425:     &   0.036264215180962 \\

2426: \hline\hline

2427: \end{tabular}

2428: \end{center}

2429: \caption{\label{tab:SAPampl} 

2430: Estimates of the critical SAP amplitude ${\cal P}(x_c)$.}

2431: \end{table}

2432:  

2433: \noindent

2434: Indeed, the observation that $a_1$ and $a_2$ vanish follows from the assumed

2435: asymptotic behaviour of $p_m$.

2436: In Table~\ref{tab:SAPampl} we have listed estimates for  ${\cal P}(x_c)$ obtained

2437: using various values of $M$ and the cut-off $K$ in the asymptotic form. From this

2438: we confidently estimate that ${\cal P}(x_c)=0.0362642151808(2)$.

2439: 

2440: \begin{figure}[htbp]

2441:   \centering

2442:   \includegraphics[scale=1.0]{Sap1_amplpred}

2443:   \caption{\label{fig:Sap1A} Estimates of the amplitude ratio $A_k^{(1)}/A_{k+1}$ for 

2444:   self-avoiding polygons with one puncture of arbitrary size.}

2445: \end{figure}

2446: 

2447: In figure~\ref{fig:Sap1A} we have plotted estimates of the amplitude ratio 

2448: $A_k^{(1)}/A_{k+1}$ for self-avoiding polygons with one puncture of arbitrary size.

2449: In the leftmost panel we look at the ratio  $A_0^{(1)}/A_{1}$ using different

2450: cut-offs $K$. The rightmost  panel shows the ratio $A_k^{(1)}/A_{k+1}$ for 

2451: area moments up to $k=9$ using the cut-off $K=10$. The straight line corresponds

2452: to the expected value ${\cal P}(x_c)$, using the estimate for this quantity obtained above. 

2453: The estimates in the leftmost panel show

2454: some variation when plotted against $1/M$, but in the limit $M\to \infty$

2455: the estimates appear to converge to the expected value (if the trend holds). As for minimally 

2456: punctured SAPs we see an ever closer agreement as $K$ is increased, again indicating 

2457: that the assumed asymptotic form is reasonable. Obviously, as seen in the rightmost panel, 

2458: the estimates for high area moments are not as close to the expected value. However,

2459: given the trend in these estimates we are confident in stating that our

2460: numerical analysis is consistent with the results of Theorem~\ref{theo:arb}.

2461: The agreement is particularly impressive bearing in mind that we analyse

2462: series with just 42 terms.

2463: 

2464: \begin{figure}[htbp]

2465:   \centering

2466:   \includegraphics[scale=0.95]{Sap2_amplpred}

2467:   \caption{\label{fig:Sap2A} Estimates of the amplitude ratio $A_k^{(2)}/A_{k+2}$ 

2468:   for self-avoiding polygons with two punctures of arbitrary size.}

2469: \end{figure}

2470: 

2471: Finally in figure~\ref{fig:Sap2A} we have plotted our estimates of the amplitude ratio 

2472: $A_k^{(2)}/A_{k+2}$ for self-avoiding polygons with two punctures of arbitrary size.

2473: The straight line corresponds to the expected value ${\cal P}(x_c)^2/2$. Again

2474: all estimates are consistent with the results of Theorem~\ref{theo:arb}, though

2475: the numerical agreement is less convincing, but then again the series have only

2476: 38 terms.

2477: 

2478: 

2479: \section{Conclusion}

2480: 

2481: We have rigorously analysed the effect of punctures on

2482: the area law of polygon models. In particular we obtained

2483: expressions for the leading amplitudes of the area moments 

2484: for punctured polygons in terms of the amplitudes for unpunctured

2485: polygons (see Theorems~\ref{theo:bounded} and~\ref{theo:arb}). For staircase

2486: polygons this led to exact formulas for the amplitudes.

2487: For self-avoiding polygons the formulas rely on an assumption about

2488: the asymptotic behaviour of SAPs without punctures. They contain constants not 

2489: known exactly but estimated numerically to a very high degree of accuracy.

2490: Our analysis also led to conjectures

2491: about the scaling behaviour of these models (see Conjecture~\ref{con1}).

2492: A proof of these conjectures is an open and difficult problem. Further

2493: numerical support of these conjectures might follow from an analysis

2494: of critical perimeter moments along the lines of \cite{RJG04}.

2495: 

2496: The expressions for the amplitudes were thoroughly checked

2497: numerically. For staircase polygons with up to 5 minimal punctures and

2498: staircase polygons with one or two punctures of fixed size

2499: we used our series expansions to find the {\em exact} generating

2500: functions for area moments to order 10. Naturally, in all these cases the

2501: leading amplitude agrees with the proved formulas.  Interestingly

2502: we find that the amplitude of the correction term is proportional

2503: to the corresponding leading amplitude with one less puncture.

2504: For staircase polygons with one and two punctures of arbitrary size

2505: a careful asymptotic analysis of the series for the area moments

2506: yielded very accurate estimates for the amplitudes, again confirming

2507: the exact formulas. Finally, we also analysed series for self-avoiding

2508: polygons with one and two punctures (minimal as well as arbitrary).

2509: In this case the numerical evidence is not quite as convincing, but

2510: we did find that the numerical estimates agree with the exact formulas

2511: to at least 3--4 significant digits.

2512: 

2513: The numerical analysis also yielded explicit expressions for correction 

2514: terms, see Eqs.~(\ref{eq:StMB}), (\ref{eq:S1C}) and subsection \ref{sec:stair_fix}.

2515: These correspond to contributions from puncture-boundary

2516: interactions and from puncture-puncture interactions. It would

2517: be interesting, but difficult to give a combinatorial proof of these results.

2518: The difficulty of any  combinatorial proof becomes clearer when

2519: considering the recent closed form solution of a closely related model

2520: of punctured staircase polygons. In \cite{IJ07} one of us (IJ) considered

2521: a model of punctured staircase polygons in which the internal polygon

2522: is rotated by 90 degrees with respect to the outer polygon. 

2523: The proofs in this paper never consider such restrictions on the 

2524: placement of the internal polygon, that is, the internal polygon can 

2525: be placed in any way one pleases. The results therefore carry over 

2526: unaltered to the problem of rotated punctured staircase polygons. 

2527: Interestingly, this means that the leading asymptotic form of

2528: the coefficients is exactly the same for both models, any differences arising 

2529: only from the correction terms. The dominant correction term to the 

2530: half-perimeter generating function for punctured staircase polygons  

2531: $\PGf(x)$ is $\propto  \log(1-4x)/\sqrt{1-4x}$ \cite{GJ06b}.

2532: From the exact closed form solution to the half-perimeter generating function 

2533: for rotated punctured staircase polygons  $\PPR(x)$ it follows that

2534: the first correction term is  $\propto (1-4x)^{-3/4}$ \cite{IJ07}. 

2535: These differences  indicate that combinatorial arguments for a

2536: proof of sub-dominant behaviour must be quite subtle!

2537: 

2538: In this paper, we discussed models of punctured polygons,

2539: where punctures are of the same type as the outer polygon.

2540: More generally, punctures may be built from different

2541: polygon classes. In that situation, Theorem 1 holds, with the obvious

2542: modification, for any collection of polygon models as punctures. 

2543: Theorem 2 holds, with the obvious modification, for any collection 

2544: of polygon models as punctures, if the corresponding half-perimeter generating 

2545: functions are finite at the radius of convergence $x_c$ of the half-perimeter generating

2546: function of the outer polygon. This includes the model considered in 

2547: \cite{IJ07} as a special case. We remark that the critical half-perimeter

2548: generating function of the outer polygon may be infinite.

2549: 

2550: Our analytical results are based on the observation that 

2551: puncture counting can be done using polygon area estimates, 

2552: polygon boundary contributions being asymptotically

2553: negligible. In particular, effects of self-avoidance 

2554: do not influence the results. This phenomenon is also

2555: expected to hold in higher dimension.  

2556: Consider so-called polycubes, the generalisation of

2557: polyominoes to three dimensions. Polycubes have been

2558: enumerated by volume up to 18 cubes, see \cite{AB06}. 

2559: Three-dimensional vesicles \cite{W93} are a subclass

2560: of polycubes, having no interior holes. Let a class of three-dimensional

2561: vesicles be given, counted by surface area, with a bounded number 

2562: of vesicular holes. Assume that the asymptotic behaviour

2563: of the volume moments is known for the model without

2564: vesicular holes. If $0<\phi<1$ and the critical surface

2565: area generating function is finite, then our method of proof 

2566: can be adapted to describe the volume law of three-dimensional 

2567: vesicles with vesicular holes. For the full model of closed

2568: self-avoiding orientable surfaces of genus zero on the cubic 

2569: lattice, however, there is numerical evidence that $\phi=1$, 

2570: see e.g.~the review in \cite{Vanderzande}.

2571: 

2572: For models with minimal punctures, the number $p_m^{\Box(1,0)}$ 

2573: in Eq.~(\ref{form:minprk}) also counts the number of 

2574: polygons winding around a fixed point 

2575: of the dual lattice. This problem and its generalisation 

2576: to $M>1$ mutually avoiding self-avoiding polygons has been 

2577: considered previously by Cardy \cite{C00}. It would be

2578: interesting to consider whether this generalisation

2579: can also be treated using the above methods. This

2580: involves the analysis of polygon models satisfying 

2581: $\theta=0$. If $\theta<0$, interaction terms are generally

2582: not asymptotically negligible, so that the above

2583: analysis does not yield asymptotically exact estimates. 

2584: 

2585: A major open question is the problem of self-avoiding polygons

2586: with an unbounded number of punctures. Here, Theorem~\ref{theo:arb}

2587: yields an upper bound.

2588: Let $\QGf_{k}(x)$ denote the $k^{th}$ area moment generating function

2589: for a model of punctured polygons with an arbitrary number of punctures.

2590: $\QGf_{k}(x)$ is a formal power series, usually with zero radius of convergence.

2591: The (asymptotically exact) upper bound for models with $r$ 

2592: punctures yields an upper estimate for the number of punctured 

2593: polygons with an arbitrary number of punctures. It is

2594: %

2595: \begin{displaymath}

2596: \QGf_{k}(x) \ll \sum_{r=0}^\infty \frac{\PGf_{k+r}(x)}{r!}(\PGf_{0}(x))^r,

2597: \end{displaymath}

2598: %

2599: where $\ll$ denotes coefficientwise majorisation. Let $\widetilde \QGf_{k}(x)$ 

2600: denote the  $k^{th}$ factorial area moment generating function

2601: for punctured polygons with an arbitrary number of punctures. 

2602: It can be shown that an upper bound is given by

2603: %

2604: \begin{displaymath}

2605: \widetilde \QGf_{k}(x) \ll \sum_{r=0}^\infty \frac{\widetilde \PGf_{k+r}(x)}{r!}

2606: (\widetilde \PGf_{0}(x))^r=\left.\frac{{\rm d}^k}{{\rm d}q^k} {\cal P}(x,q)

2607: \right|_{q={1+\cal P}(x,1)},

2608: \end{displaymath}

2609: %

2610: where $\widetilde \PGf_{k}(x)$ is the $k^{th}$ factorial moment 

2611: generating function, and ${\cal P}(x,q)$ is the perimeter and area 

2612: generating function of the model without punctures.  In particular,

2613: the half-perimeter generating function $\QGf_{0}(x)=\widetilde 

2614: \QGf_{0}(x)$ is majorised by

2615: %

2616: \begin{displaymath}

2617: \QGf_{0}(x) \ll {\cal P}(x,1+{\cal P}(x,1)).

2618: \end{displaymath}

2619: %

2620: 

2621: Another problem touched upon in this article is punctured

2622: polygons in the fixed area ensemble. We gave a (non-rigorous)

2623: argument for values of the critical

2624: exponent in the branched polymer phase from the crossover 

2625: behaviour of the tentative scaling function, thereby confirming 

2626: previous results \cite{JvR92}.

2627: In that phase, boundary effects are indeed crucial, such that 

2628: our methods of deriving limit distributions cannot be applied in this

2629: situation. On the other hand, area laws are expected to be

2630: of Gaussian type, as is usually the case away from

2631: phase transition points. The same phenomenon is expected

2632: to occur in the fixed perimeter ensembles for $q\ne1$.

2633: 

2634: 

2635: \section*{E-mail or WWW retrieval of series}

2636: 

2637: The series for the generating functions studied in this paper 

2638: can be obtained via e-mail by sending a request to 

2639: I.Jensen@ms.unimelb.edu.au or via the world wide web on the URL

2640: http://www.ms.unimelb.edu.au/\~{ }iwan/ by following the instructions.

2641: 

2642: 

2643: \section*{Acknowledgements}

2644: 

2645: The calculations presented in this paper would not have been possible

2646: without a generous grant of computer time on the server cluster of the

2647: Australian Partnership for Advanced Computing (APAC). We also used

2648: the computational resources of the Victorian Partnership for Advanced 

2649: Computing (VPAC). 

2650: CR and AJG would like to acknowledge support by the German Research Council

2651: (Deutsche Forschungsgemeinschaft) within the CRC701.

2652: IJ and AJG gratefully acknowledge financial support from the Australian Research Council.

2653: 

2654: 

2655: \section*{Appendix}

2656: 

2657: We analyse the asymptotic growth of a Cauchy product of sequences

2658: in terms of the asymptotic growth of its constituting sequences.

2659: The following lemma is an extension of \cite[Thm~2]{B74} to the case

2660: of generating functions with equal radii of convergence. Its proof relies

2661: on Lebesgue's dominated convergence theorem \cite[Thm~1.34]{R87}, which 

2662: states conditions under which an exchange of limit and sum is allowed:  

2663: With $n\in\mathbb N$ and $k\in\mathbb N_0$ let real numbers 

2664: $a_{n,k}$ be given. Assume that $\lim_{n\to\infty}a_{n,k}=:a_k\in\mathbb R$ 

2665: for all $k$ and that for all $k$ there is a bound $|a_{n,k}|\le b_k$ 

2666: uniformly in $n\in\mathbb N$. Assume that $\sum_{k\ge0}  b_k\in

2667: \mathbb R$. Then $\lim_{n\to\infty} \sum_{k\ge0} a_{n,k} =

2668: \sum_{k\ge0} a_k\in\mathbb R$. 

2669: 

2670: \begin{lem}

2671: Let two sequences $(f_n)_{n\in\mathbb N_0}$ and $(g_n)_{n\in\mathbb N_0}$

2672: of real numbers be given, with generating functions $f(x)=

2673: \sum_{n\ge0}f_n x^n$ and $g(x)=\sum_{n\ge0}g_n x^n$. Assume that both 

2674: generating functions have the same positive finite radius of convergence 

2675: $x_c$, $0<x_c<\infty$. Assume that the sequences $(f_n)$ and $(g_n)$ 

2676: satisfy asymptotically

2677: %

2678: \begin{displaymath}

2679: f_n \sim Ax_c^{-n}n^{\gamma-1},\qquad g_n \sim Bx_c^{-n}n^{\delta-1} \qquad

2680: (n\to\infty),

2681: \end{displaymath}

2682: %

2683: for nonzero numbers $A\ne0$, $B\ne0$, and for real constants $\gamma,\delta$

2684: satisfying $\delta<0$ and $\gamma>\delta+1$. Assume that $g(x_c):=

2685: \lim_{x\nearrow x_c} g(x)\ne0$. Then, the Cauchy product of $(f_n)$ and $(g_n)$  

2686: satisfies

2687: %

2688: \begin{displaymath}

2689: \sum_{k=0}^n f_{n-k} g_k = [x^n]f(x)g(x)\sim

2690: g(x_c) f_n \qquad (n\to\infty).

2691: \end{displaymath}

2692: %

2693: \end{lem}

2694: 

2695: \begin{proof}

2696: Let $f_{n}:=0$ for $n<0$ and define for $n\in\mathbb N$ and for $k\in\mathbb N_0$

2697: %

2698: \begin{displaymath}

2699: a_{n,k}:=\frac{f_{n-k}}{Ax_c^{-n}n^{\gamma-1}}g_k.

2700: \end{displaymath}

2701: %

2702: Below, we derive a bound on $|a_{n,k}|$ uniformly in $n\in\mathbb N$ and

2703: summable in $k\in\mathbb N_0$. Then, Lebesgue's dominated convergence theorem

2704: can be applied to $a_{n,k}$. Since $g(x_c)\ne0$, this 

2705: yields the statement of the lemma. 

2706: 

2707: Note first that the assumption on the asymptotic 

2708: behaviour of $f_n$ implies the existence

2709: of a constant $n_0\in\mathbb N$ such that for all $n\ge n_0+k$ we have the

2710: estimate

2711: %

2712: \begin{displaymath}

2713: \left|\frac{f_{n-k}}{Ax_c^{-n}n^{\gamma-1}}\right| 

2714: \le 2 x_c^k \left(1-\frac{k}{n}\right)^{\gamma-1}.

2715: \end{displaymath}

2716: %

2717: Fix such $n_0$. We distinguish the three cases $n< k$, $k\le n<n_0+k$, and $n\ge n_0+k$.

2718: If $n< k$, we clearly have $a_{n,k}=0$. For $n\ge n_0+k$, 

2719: the above estimate yields

2720: %

2721: \begin{eqnarray*}

2722: \left|a_{n,k}\right|&&=\left|\frac{f_{n-k}}{Ax_c^{-n}n^{\gamma-1}}g_k\right|

2723: \le 2\left(1-\frac{k}{n}\right)^{\gamma-1}x_c^k \left|g_k\right|.

2724: \end{eqnarray*}

2725: %

2726: We will first consider the case $\gamma-1<0$. If $n\ge n_0+k$, we have

2727: $1/(1-k/n)\le1+k/n_0$. This implies

2728: %

2729: \begin{eqnarray*}

2730: \left|a_{n,k}\right| &\le 2^{2-\gamma_1} k^{1-\gamma}x_c^k\left|g_k\right|=:b_k^{(1)}.

2731: \end{eqnarray*}

2732: %

2733: If $k\le n<n_0+k$, we estimate similarly

2734: %

2735: \begin{eqnarray*}

2736: &\hspace{-1.5cm}\left|a_{n,k}\right|=\left|\frac{f_{n-k}}{Ax_c^{-n}n^{\gamma-1}}g_k\right|

2737: \le |A|^{-1}\max\{1,x_c^{n_0}\} \max_{m<n_0}\{|f_m|\} (n_0+k)^{1-\gamma}x_c^k |g_k|\\

2738: &\le |A|^{-1}\max\{1,x_c^{n_0}\}\max_{m<n_0}\{|f_m|\} \cdot

2739: \left\{\begin{array}{cc}n_0^{1-\gamma}|g_0|, & k=0\\

2740: (2n_0)^{1-\gamma}x_c^k k^{1-\gamma}|g_k|, & k\ne0

2741: \end{array}

2742: \right\}=: 

2743: b_k^{(2)}.

2744: \end{eqnarray*}

2745: %

2746: Define $b_k:=b_k^{(1)}+b_k^{(2)}$ for $k\in\mathbb N_0$. 

2747: Then $|a_{n,k}|\le b_k$ uniformly in $n\in\mathbb N$, 

2748: and $\sum_{k\ge0}b_k<\infty$.

2749: Summability follows from

2750: %

2751: \begin{displaymath}

2752: \sum_{k=0}^\infty x_c^k k^{1-\gamma}|g_k|<\infty,

2753: \end{displaymath}

2754: %

2755: since by assumption $x_c^k k^{1-\gamma}|g_k|\sim |B| k^{\delta-\gamma}$ as $k\to\infty$,

2756: where $\gamma-\delta>1$.

2757: 

2758: If $\gamma-1\ge0$, uniform estimates are obtained along the same lines.

2759: In that situation, the factors $(1-k/n)^{\gamma-1}$ and $(n_0+k)^{1-\gamma}$

2760: may be replaced by unity, resulting in simpler uniform bounds involving 

2761: $\sum_{k\ge0} x_c^k |g_k|<\infty$.

2762: \end{proof}

2763: 

2764: 

2765: \begin{thebibliography}{10}

2766: 

2767: \bibitem{MSbook}

2768: Madras N and Slade G 1993 {\it The Self-Avoiding Walk\/} (Boston: Birkh\"auser)

2769: 

2770: \bibitem{HughesV1}

2771: Hughes B~D 1995 {\it Random Walks and Random Environments, Vol {I} Random

2772:   Walks\/} (Oxford: Clarendon)

2773: 

2774: \bibitem{Vanderzande}

2775: Vanderzande C 1998

2776: {\em Lattice Models of Polymers} Cambridge Lecture Notes in Physics 11

2777:  (Cambridge: Cambridge University Press)

2778: 

2779: \bibitem{deGennes} de Gennes P G 1979 {\it Scaling Concepts in Polymer Physics} 

2780: (Ithaca: Cornell University Press)

2781: 

2782: \bibitem{LSF87}

2783: Leibler S, Singh R~R~P and Fisher M~E 1987 Thermodynamic behavior of

2784:   two-dimensional vesicles {\em Phys. Rev. Lett.\/} {\bf 59} 1989--1992

2785: 

2786: \bibitem{GCbook} 

2787: Gutman I and Cyvin S 1989 {\it Introduction to the Theory of Benzenoid Hydrocarbons}

2788: (Berlin: Springer)

2789: 

2790: \bibitem{VGJ02}

2791: V\"oge M, Guttmann A~J and Jensen I 2002 On the number of benzenoid hydrocarbons,

2792: {\em J. Chem. Inf. Comput. Sci.\/} {\bf 42} 456--66

2793: 

2794: \bibitem{FGW91}

2795: Fisher M~E, Guttmann A~J and Whittington S~G 1991,

2796: Two-dimensional lattice vesicles and polygons,

2797: {\em J.~Phys.~A:~Math.~Gen.} {\bf 24}, 3095--3106

2798: 

2799: \bibitem{N82}

2800: Nienhuis B 1982 Exact critical point and critical exponents of O$(n)$ models in two dimensions

2801:    {\em Phys. Rev. Letts.\/} {\bf 49} 1062--65

2802: 

2803: \bibitem{N84}

2804: Nienhuis B 1984 Critical behavior of two-dimensional spin models and charge

2805:   asymmetry in the coulomb gas {\em J. Stat. Phys.\/} {\bf 34} 731--761

2806: 

2807: \bibitem{RGJ01}

2808: Richard C, Guttmann A~J and Jensen I 2001 Scaling function and universal amplitude

2809: combinations for self-avoiding polygons {\em J.~Phys.~A:~Math.~Gen. \/} {\bf 34} L495--501

2810: 

2811: \bibitem{C01}

2812: Cardy J 2001

2813: Exact scaling functions for self-avoiding loops and branched polymers

2814: {\it J.~Phys.~A:~Math.~Gen.~\bf 34}  L665--L672

2815: 

2816: \bibitem{R02}

2817: Richard C 2002 Scaling behaviour of two-dimensional polygon models {\em J.

2818:   Stat. Phys.\/} {\bf 108} 459--493

2819: 

2820: \bibitem{RJG03}

2821: Richard C, Jensen I and Guttmann A~J 2003 Scaling function for self-avoiding

2822:   polygons in {\em Proceedings of the International Congress on Theoretical

2823:   Physics TH2002 (Paris), Supplement\/} (eds Iagolnitzer D, Rivasseau V and

2824:   Zinn-Justin J) (Basel: Birkh\"auser) 267--277.

2825: 

2826: \bibitem{RJG04}

2827: Richard C, Jensen I and Guttmann A~J 2004 Scaling function for self-avoiding

2828:   polygons revisited {\em J. Stat. Mech.: Th. Exp.\/}  P08007

2829: 

2830: \bibitem{GJWE00}

2831: Guttmann A~J, Jensen I, Wong L~H and Enting I~G 2000 Punctured polygons and

2832:   polyominoes on the square lattice {\em J. Phys. A: Math. Gen.\/} {\bf 33}

2833:   1735--1764

2834: 

2835: \bibitem{JvRW89}

2836: Janse~van Rensburg E~J and Whittington S~G 1989 Self-avoiding surfaces 

2837: {\em J. Phys. A: Math. Gen.\/} {\bf 22} 4939--4958

2838: 

2839: \bibitem{GJO01}

2840: Guttmann A~J, Jensen I and Owczarek A~L 2001 Polygonal polyominoes on the square

2841:   lattice {\em J. Phys. A: Math. Gen.\/} {\bf 34} 3721--3733

2842: 

2843: \bibitem{J99}

2844: Janse van Rensburg E~J 1999

2845: Models of composite polygons

2846:  {\em J. Phys. A: Math. Gen.\/} {\bf 32} 4351--4372

2847: 

2848: \bibitem{BH75}

2849: Bleistein N and Handelsman Richard~A 1986

2850: {\em Asymptotic Expansions of Integrals} 2nd ed (New York: Dover)

2851: 

2852: \bibitem{JvR92}

2853: Janse~van Rensburg E~J 1992 Surfaces in the hypercubic lattice 

2854: {\em J. Phys. A: Math. Gen.\/} {\bf 25} 3529--3547

2855: 

2856: \bibitem{GJ06b}

2857: Guttmann A~J and Jensen I 2006 The perimeter generating function of punctured

2858:   staircase polygons {\em J. Phys. A: Math. Gen.\/} {\bf 39} 3871--3882

2859: 

2860: \bibitem{OP99a}

2861: Prellberg T and Owczarek A 1999 On the asymptotics of the finite-perimeter

2862:   partition function of two-dimensional lattice vesicles {\em Commun. Math.

2863:   Phys.\/} {\bf 201} 493--505

2864: 

2865: \bibitem{J00}

2866: Janse van Rensburg E J 2000

2867: {\it The Statistical Mechanics of Interacting Walks, Polygons, 

2868: Animals and Vesicles}

2869: Oxford Lecture Series in Mathematics and its Applications

2870: {\bf 18} (Oxford: Oxford University Press)

2871: 

2872: \bibitem{FS06}

2873: Flajolet P and Sedgewick R 2008 \textit{Analytic Combinatorics} in press

2874: (Cambridge: Cambridge University Press)

2875: 

2876: \bibitem{BI03}

2877: Brydges D~C and Imbrie J~Z 2003 End-to-end distance from the {G}reen's function

2878:   for a hierarchical self-avoiding walk in four dimensions {\em Commun. Math.

2879:   Phys.\/} {\bf 239} 523--547

2880: 

2881: \bibitem{Pr94}

2882: Prellberg T 1994 Uniform $q$-series asymptotics for staircase polygons

2883:    {\em J. Phys. A: Math. Gen.\/} {\bf 28} 1289--1304

2884: 

2885: \bibitem{L84}

2886: Louchard G 1984 The Brownian excursion area: a numerical analysis

2887: {\em Comput. Math. Appl. \bf 10} 413--417;

2888: Louchard G 1986

2889: Erratum: ``The Brownian excursion area: a numerical analysis"  

2890: {\em Comput. Math. Appl. \bf 12} 375

2891: 

2892: \bibitem{T91}

2893: Tak\'acs L 1991 On a probability problem connected with railway

2894: traffic {\em J.~Appl.~Math.~Stoch.~Anal. \bf 4} 1--27

2895: 

2896: \bibitem{FL01}

2897: Flajolet P and Louchard G 2001 Analytic variations on the Airy

2898: distribution {\em Algorithmica \bf 31} 361--377

2899: 

2900: \bibitem{C74}

2901: Chung K L 1974

2902: {\it A Course in Probability Theory}

2903: (New York: Academic Press) 2nd ed

2904: 

2905: \bibitem{F99}

2906: Flajolet P 1999 Singularity analysis and asymptotics of Bernoulli sums

2907: {\em Theoret.~Comput.~Sci.} {\bf 215} 371--381

2908: 

2909: \bibitem{R05}

2910: Richard C 2008 On $q$-functional equations and excursion moments

2911: {\em Discrete Math.}, in press; {\tt math.CO/0503198}

2912: 

2913: \bibitem{R06}

2914: Richard C 2006 Staircase polygons: moments of diagonal lengths and column

2915:   heights {\em J. Phys.: Conf. Ser.} {\bf 42} 239--257

2916: 

2917: \bibitem{BM96} 

2918: Bousquet-M\'elou M 1996 

2919: A method for the enumeration of various classes of column-convex polygons

2920: {\em Discrete~Math.~\bf 154} 1--25

2921: 

2922: \bibitem{BOP93}

2923: Brak R, Owczarek A L and Prellberg T 1993

2924: A scaling theory of the collapse transition in geometric cluster

2925: models of polymers and vesicles

2926: {\em J. Phys. A: Math. Gen.\/} {\bf 26} 4565--4579

2927: 

2928: \bibitem{RG01}

2929: Richard C and Guttmann A~J 2001 $q$-linear approximants: scaling functions for

2930:   polygon models {\em J. Phys. A: Math. Gen.\/} {\bf 34} 4783--4796

2931: 

2932: \bibitem{IJ03}

2933: Jensen I 2003 A parallel algorithm for the enumeration of self-avoiding

2934:   polygons on the square lattice {\em J. Phys. A: Math. Gen.\/} {\bf 36}

2935:   5731--5745

2936: 

2937: \bibitem{IJ00} Jensen I 2000 Size and area of square lattice polygons 

2938: {\em J. Phys. A: Math. Gen.\/} {\bf 33}  3533--3543

2939: 

2940: \bibitem{AJG89a}

2941: Guttmann A~J 1989 Asymptotic analysis of power-series expansions in {\em Phase

2942:   Transitions and Critical Phenomena\/}  vol.~13 (eds Domb C and Lebowitz J L) (New

2943:   York: Academic) 1--234

2944: 

2945: \bibitem{IJ06a}

2946: Jensen I 2006 Honeycomb lattice polygons and walks as a test of series analysis

2947:   techniques {\em J. Phys.: Conf. Ser.} {\bf 42} 163--178.

2948: 

2949: \bibitem{JG99}

2950: Jensen I and Guttmann A~J 1999 Self-avoiding polygons

2951: on the square lattice {\em J. Phys. A: Math. Gen.\/} {\bf 32} 4867--4876

2952: 

2953: \bibitem{IJ07}

2954:  Jensen I 2006 Exact perimeter generating function for a model of punctured staircase polygons

2955: Preprint: {\tt  cond-mat/0610605}

2956: 

2957: \bibitem{AB06}

2958: Aleksandrowicz G and Barequet G 2006 

2959: Counting $d$-dimensional polycubes and nonrectangular planar polyominoes

2960: in: {\it Proc. 12th Ann. Int. Computing and Combinatorics Conf.   (COCOON), Taipei, Taiwan}  

2961: Springer Lecture Notes in Computer Science {\bf 4112} 418--427

2962: 

2963: \bibitem{W93}

2964: Whittington S~G 1993

2965: Statistical mechanics of three-dimensional vesicles

2966: {\it J.~Math.~Chem. \bf 14} 103--100

2967: 

2968: \bibitem{C00} 

2969: Cardy J 2000 Linking numbers for self-avoiding loops and percolation: 

2970:  application to the spin quantum hall transition {\em Phys.~Rev.~Lett.\/} 

2971:  {\bf 84} 3507--3510

2972: 

2973: \bibitem{B74}

2974: Bender E~A 1974

2975: Asymptotic methods in enumeration

2976: {\em SIAM Rev.  \bf 16}  485--515;

2977: Errata {\em SIAM Rev.  \bf 18}  (1976) 292

2978: 

2979: \bibitem{R87}

2980: Rudin W 1987 {\em Real and Complex Analysis} (New York: McGraw-Hill)

2981: 

2982: \end{thebibliography}

2983: 

2984: 

2985: \end{document}

2986: