1: \documentclass[12pt]{article}

2: \usepackage{amsfonts}

3: \usepackage{amssymb}

4: \usepackage[dvips]{graphics}

5: \pagestyle{myheadings}

6: \markright{bolina\\ \\ \\  \\ \\ \\ \\ \\ \\ }

7: %

8: \newcommand{\Rl}{\mathbb{R}}

9: \newcommand{\Nl}{\mathbb{N}}

10: \newcommand{\Ir}{\mathbb{Z}}

11: \newcommand{\Cx}{\mathbb{C}}

12: \newcommand{\A}{\mathcal{A}}

13: \newcommand{\HH}{\mathcal{H}}

14: \newcommand{\N}{\mathcal{N}}

15: \newcommand{\glb}{{\rm glb}}

16: %

17: \newtheorem{theorem}{Theorem}[section]

18: \renewcommand{\thetheorem}{\thesection.\arabic{theorem}}          

19: \newtheorem{lemma}[theorem]{Lemma}              

20: \newtheorem{proposition}[theorem]{Proposition}

21: \newtheorem{corollary}[theorem]{Corollary}

22: \newtheorem{definition}[theorem]{Definition}

23: \newtheorem{conjecture}[theorem]{Conjecture}

24: %

25: \newcommand{\rem}[1]{{\bf Remark:}~}

26: %

27: \newcommand{\condmat}[1]{archived as {\tt cond-mat/#1}}

28: \newcommand{\Section}[1]{\setcounter{equation}{0}\section{#1}}

29: \renewcommand{\theequation}{\thesection.\arabic{equation}}

30: \newcommand{\eq}[1]{(\ref{#1})}

31: %

32: \newenvironment{proof}{\noindent {\bf Proof: }}{\QED\medskip}

33: \def\QED{{\hspace*{\fill}{\vrule height .5ex width 1ex }\quad} 

34:     \vskip 0pt plus20pt}

35: %

36: \newcommand{\pprime}{{\prime\prime}}

37: \def\idty{{\mathchoice {\rm 1\mskip-4mu l} {\rm 1\mskip-4mu l} % 

38: {\rm 1\mskip-4.5mu l} {\rm 1\mskip-5mu l}}}

39: %

40: \newcommand{\be}{\begin{equation}}

41: \newcommand{\ee}{\end{equation}}

42: \newcommand{\bea}{\begin{eqnarray}}

43: \newcommand{\eea}{\end{eqnarray}}

44: \newcommand{\beann}{\begin{eqnarray*}}

45: \newcommand{\eeann}{\end{eqnarray*}}

46: \newcommand{\bi}{\begin{itemize}}

47: \newcommand{\ei}{\end{itemize}}

48: %

49: \newcommand{\heisen}{{\bf S}_x\cdot{\bf S}_{x+1}}

50: \newcommand{\ket}[1]{\left\vert #1\right\rangle}

51: \newcommand{\bra}[1]{\left\langle #1\right\vert}

52: \newcommand{\wslim}{{\rm w}^*\mbox{-}\lim}

53: \newcommand{\Prob}{{\rm Prob}}

54: 

55: 

56: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

57: % Shannon's macros

58: %%%%%%%%%%%%%%%%%%%%%%%%

59: \DeclareMathAlphabet{\mathol}{OT1}{cmr}{l}{ol}

60: 

61: \newcommand{\up}{\ket{\uparrow}}

62: \newcommand{\down}{\ket{\downarrow}}

63: \newcommand{\uu}{\ket{\uparrow \uparrow}}

64: \newcommand{\ud}{\ket{\uparrow \downarrow}}

65: \newcommand{\du}{\ket{\downarrow \uparrow}}

66: \newcommand{\dd}{\ket{\downarrow \downarrow}}

67: \newcommand{\C}{\mathbb{C}}

68: \newcommand{\alphavec}{\mathbf{\alpha}}

69: \newcommand{\alphastate}{\ket{\alphavec}}

70: \newcommand{\Ac}{\mathcal{A}}

71: \newcommand{\D}{\mathcal{D}}

72: \newcommand{\Z}{\mathbb{Z}}

73: \newcommand{\ZGC}[2]{Z^{GC}(#1,#2)}

74: \newcommand{\floor}[1]{\lfloor{#1}\rfloor}

75: \newcommand{\ip}[2]{\langle{#1|#2}\rangle}

76: \newcommand{\op}[3]{\bra{#1} #2 \ket{#3}}

77: \newcommand{\dist}{dist}

78: \newcommand{\h}{\mathsf{h}}

79: \newcommand{\avg}[2]{\langle{#1}\rangle_{#2}}

80: \newcommand{\avgGC}[2]{\langle{#1}\rangle^{GC}_{#2}}

81: \newcommand{\unity}{1\hskip -3pt \rm{I}}

82: \newcommand{\supp}{{\rm supp}}

83: \newcommand{\Num}{\mathsf{N}}

84: \newcommand{\navg}{\avgGC{\Num}{\Sigma,\mu}}

85: \newcommand{\Sc}{\mathcal{S}}

86: 

87: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

88: 

89: 

90: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

91: 

92: \begin{document}

93: %{\baselineskip=10pt \thispagestyle{empty} {{\small Preprint UC Davis Math

94: %1999-24}

95: %{\tt cond-mat/0111242n} and {\tt mp\_arc 99-nnn} 

96: 

97: \vspace{20pt}

98: 

99: \title{\vspace*{-.35in}

100: }

101: \author{{\bf Oscar Bolina}\\[6pt]

102: %EndAName

103: Instituto de F\'{\i}sica\\

104: Universidade de S\~ao Paulo\\

105: Caixa Postal 66318 \\

106: S\~ao Paulo 05315-970 Brasil  \\

107: {\bf E-mail:} bolina@if.usp.br

108: }

109: \title{\vspace{-1in}

110: {

111: \bf 

112: The gambler's ruin problem in path representation form  

113: }}

114: \date{}

115: \maketitle

116: \begin{abstract}

117: \noindent 

118: We analyze the one-dimensional random walk of a particle on 

119: the right-half real line. The particles starts at $x=k$, 

120: for $k >0$, and ends up at the origin. We solve for the 

121: probabilities of absorption at the origin by means of a

122: geometric representation of this random walk in terms 

123: of paths on a two-dimensional lattice.

124: \vskip .2 truecm

125: \noindent

126: {\bf Key words:} Random Walk.

127: \newline

128: {\bf MSC numbers:} 82B41

129: \end{abstract}

130: \section{Introduction}

131: \noindent 

132: We consider the classical one-dimensional random walk of a particle

133: on the right-half real line. We assume that the particle is 

134: initially at position $x=k$, $k > 0$, and moves to the right with 

135: probability {\it p} or to the left with probability $1-p$. We 

136: consider that the particle is absorbed at the origin without 

137: fixing the number of steps needed to get there. We calculate 

138: the probability $P(x=k)$ that the particles end up at the origin, 

139: given that it starts at $x=k$, by means of a geometric 

140: representation of this random walk in terms of paths on a 

141: two-dimensional lattice.

142: \newline

143: It is well-known that the probability $P(x=k)$ is the $k^{th}$ power 

144: of $P(x=1)$:

145: \be\label{xi}

146: P(x=k)=\left ( P(x=1) \right )^{k}.

147: \ee

148: Because of this, we are not really going to {\it prove} the results 

149: that lead to (\ref{xi}) (the proofs are quite obvious in our

150: context). Instead, we prefer to use our geometric 

151: representation as a "visual" proof of these results.

152: \newpage

153: \section{The probabilities of absorption}

154: A representation of this random walk in terms 

155: of path on a two-dimensional lattice is given in Fig. \ref{fig1} 

156: for an arbitrary initial position $x=k$ ($k > 0$) of the 

157: particle. In this representation, each horizontal lattice 

158: bond represents a step of the particle to the right, 

159: while each vertical bond represents a step to the left. 

160: \newline

161: Let {\it n} denote the total number of steps of the particle 

162: to the right until it is absorbed at the origin. (The number of 

163: steps to the left will then be $n+k$.)  

164: \newline

165: A path of length {\it n} --- or simply a path --- of the particle 

166: on this lattice is the set of horizontal and vertical bonds from 

167: the origin to the point of coordinates $(n, n+k)$ on the line 

168: {\it L}, as shown.

169: %\newline

170: %The variable {\it m} is not an independent

171: %variable, but depends on {\it n} and {\it k} by $m=n+k$. 

172: \newline

173: Let $C_{k}(n)$ be the {\it total} number of paths of length {\it n} 

174: from the origin to point $(n, n+k)$. A simple combinatorics analysis 

175: leads to the following expression for $C_{k}(n)$:

176: \be\label{fi}

177: C_{k}(n)=k~ \frac{(n+m-1)!}{{n!}~{m!}}=k~\frac{(2n+k-1)!}{{n!}(n+k)!}

178: =\frac{k}{(2n+k)} \left (

179: \begin{array}{cc}

180: 2n+k \\

181: n

182: \end{array}

183: \right ).

184: \ee

185: The probability that 

186: %This number is related to the probabilities of absorption of the 

187: %particles at the origin. 

188: %A simple combinatorics analysis leads to the following expression for

189: the particle arrives at the origin given the initial

190: condition  that it starts at position $x=k$ is given by

191: \be\label{pf}

192: P(x=k)=\sum_{n=0}^{\infty}C_{k}(n) p^{n}(1-p)^{n+k}.

193: \ee

194: We show next how to carry out the summation in (\ref{pf}) using the

195: geometry of the lattice paths. We will need to solve explicitly for the

196: first three values of $k$, $k=1, 2, 3$, before arriving at the general 

197: solution (\ref{xi}) by induction.  

198: \vskip .3 cm

199: \noindent

200: \section{The case $k=1$} 

201: \vskip .3 cm

202: \noindent

203: The expression (\ref{fi}) for $C_{k=1}(n)=C(n)$ in the particular case 

204: the particle starts at $k=1$ is the $n^{th}$ Catalan number 

205: \be\label{11}

206: C(n)=\frac{(2n)!}{{n!}~{(n+1)!}}=\frac{1}{n+1} \left (

207: \begin{array}{cc}

208: 2n \\

209: n

210: \end{array}

211: \right ).

212: \ee

213: The Catalan numbers have some geometric properties associated 

214: with paths on the lattice that we now explore. The first result

215: is the following

216: \vskip .1 cm

217: \noindent

218: {\theorem The number $C(n)$ given by (\ref{11}) satisfies the following 

219: equation 

220: \be\label{phin}

221: C(n)=\sum_{\alpha=1}^{n} C(\alpha-1)~C(n-\alpha).

222: \ee

223: }

224: \begin{figure}

225: \begin{center}

226: \resizebox{!}{8 truecm}{\includegraphics{pp.eps}}

227: \vskip .2 cm

228: \parbox{13truecm}{\caption{\baselineskip=16 pt\small \label{fig1}

229: Paths on the lattice, from the origin to the point $(n, n+k)$ on the line

230: {\it L}, represent the number of steps to the right (n), and to 

231: the left (n+k) that the particle has taken in its random walk 

232: from position $x=k$ on the real axis until it is absorbed at the origin. }

233: }

234: \end{center}

235: \end{figure}

236: \begin{itemize}

237: \item[] {\bf Proof.} Fig. \ref{fig2} illustrates the theorem for

238: $n=4$. For this proof, we have translated the paths in $C(4)$ one unit to

239: the left so as to make the paths touch the line {\it L}. These paths can

240: be decomposed as follows. Starting from $n=0$, and repeating the 

241: procedure for $n=1$, $n=2$ and $n=3$, raise vertical lines 

242: from the {\it n}-axis to the line {\it L}. The vertical line 

243: at $n=\alpha$ generates two sets of paths, the set 

244: $C(\alpha)$ to the left of $n=\alpha$, and the set 

245: $C(4-\alpha)$ to the right of $n=\alpha$. These two sets 

246: of paths contribute a factor $C(\alpha)C(4-\alpha)$ 

247: to the number $C(4)$. The sum of these factors for 

248: values from zero to $3$ is $C(4)$.

249: \hfill $\blacksquare$

250: \end{itemize} 

251: \subsection{Finding $P(x=1)$}

252: \vskip .3 cm

253: \noindent

254: Now if we set

255: \be\label{sum}

256: F(z)=\sum_{\alpha=1}^{\infty}~C(\alpha-1)~z^{\alpha},

257: \ee

258: then we obtain

259: \begin{eqnarray}\label{sum2}

260: F^{2}(z)&=&\sum_{\alpha,\beta=1}^{\infty}~C(\alpha-1)C(\beta-1)~z^{\alpha 

261: +\beta}  \nonumber\\

262: &=& 

263: \sum_{n=2}^{\infty}~\sum_{\alpha=1}^{n-1}~C(\alpha-1)C(n-\alpha 

264: -1)~z^{n} \nonumber\\

265: &=& \sum_{n=2}^{\infty}~C(n-1)~z^{n}=F(z)-z, \nonumber 

266: \end{eqnarray}

267: so that (\ref{sum}) obeys the equation 

268: \be\label{1}

269: F^{2}(z)=F(z)-z.

270: \ee

271: \noindent

272: \begin{figure}

273: \begin{center}

274: \resizebox{!}{3.2 truecm}{\includegraphics{pp3.eps}}

275: \vskip .3 cm

276: \parbox{13truecm}{\caption{\baselineskip=16 pt\small\label{fig2}

277: Decomposition of $C(n)$.}

278: }

279: \end{center}

280: \end{figure}

281: Solving (\ref{1}) for {\it F} we get the solutions

282: \be

283: F(z)=\frac{1 \pm \sqrt{1-4z}}{2}.

284: \ee

285: Observe now that the summation (\ref{sum}) when $z=p(1-p)$ is (up to a

286: factor) the probability function (\ref{pf}) when $k=1$. Thus we  have

287: \be

288: P(x=1)=\frac{F(p-p^{2})}{p}=\frac{1 \pm

289: \sqrt{1-4p+4p^{2}}}{2p}=\frac{1 \pm (1-2p)}{2p}.

290: \ee

291: This last expression gives the classical results for the

292: probability that the particle will end up at the origin 

293: given that it starts at $k=1$. The solutions depend on

294: whether $p \leq  1/2$ or $p \geq 1/2$. For $p  \leq 1/2$, 

295: we get $P(x=1)=1$, and for $p  \geq 1/2$ we get 

296: \be\label{i}

297: P(x=1)=\frac{1-p}{p}.

298: \ee

299: For an interpretation of these results for other values of

300: {\it p}, see reference \cite{FM}.

301: \vskip .3 cm

302: \noindent

303: \section{The case $k=2$}

304: \vskip .3 cm

305: \noindent

306: When $k=2$, then (\ref{pf}) becomes 

307: \be\label{pf1}

308: P(x=2)=\sum_{n=0}^{\infty}C_{2}(n) p^{n}(1-p)^{n+2}.

309: \ee

310: The coefficient $C_{2}(n)$ can be written in terms of $C(n)$ on

311: account of the following result

312: \vskip .3 cm

313: \noindent

314: {\theorem The following relationship holds for the coefficients $C_{2}(n)$

315: and $C(n)$ given by (\ref{fi}) 

316: \be\label{phin1}

317: C_{2}(n)=C(n+1).

318: \ee

319: }

320: \begin{itemize}

321: \item[] {\bf Proof.} In Fig. \ref{fig3} we have drawn the paths

322: that in $C_{2}(n)$ (in

323: red), and the paths in $C(n+1)$. Now, translate the paths in blue one 

324: unit to the left.  The blue paths go over the red ones. 

325: It is a perfect match and we are done.

326: \hfill $\blacksquare$

327: \end{itemize}

328: \vskip .5 cm

329: \noindent

330: If we now substitute (\ref{phin1}) into (\ref{pf1}) we obtain,

331: after some simple calculations,

332: \[

333: P(x=2)= \sum_{n=0}^{\infty} C(n+1)p^{n}(1-p)^{n+2} 

334: =\sum_{l=1}^{\infty} C(l)p^{l-1}(1-p)^{l+1}=

335: \frac{P(x=1)}{p}-\frac{1-p}{p} 

336: \]

337: and substituting for $P(x=1)$ from (\ref{i}) we get

338: \[

339: P(x=2)=\frac{(1-p)^{2}}{p^2} 

340: \]

341: if $p \geq 1/2$,  and $P(x=2)=1$ for $p \leq 1/2$.

342: \noindent

343: \begin{figure}

344: \begin{center}

345: \resizebox{!}{5 truecm}{\includegraphics{pp1.eps}}

346: \vskip .3 cm

347: \parbox{13truecm}{\caption{\baselineskip=16 pt\small \label{fig3}

348: Relationship between $C_{2}(n)$ and the Catalan numbers.}

349: }

350: \end{center}

351: \end{figure}

352: \vskip .3 cm

353: \noindent

354: \section{The case $k=3$}

355: \vskip .3 cm

356: \noindent

357: We need this case to establish (\ref{xi}) for all values of {\it k}.  

358: \vskip .2 cm

359: \noindent

360: {\theorem

361: For $k \geq 3$ the following recurrence relation holds 

362: between the coefficient $C_{k}(n)$ and the coefficients $C_{k-1}(n+1)$ e

363: $C_{k-2}(n+1)$ 

364: \vskip .5 cm

365: \noindent

366: \be\label{phin3}

367: C_{k}(n)=C_{k-1}(n+1)-C_{k-2}(n+1).

368: \ee

369: }

370: \begin{itemize}

371: \item[] {\bf Proof.} The geometric representation of (\ref{phin3}) is

372: depicted in Fig. \ref{fig4}. To prove it, note that if we remove the paths

373: that stem from the two purple bonds on the right side of Fig. \ref{fig4} 

374: and translate the remaining set of paths one unit to the left, we get

375: exactly $C_{k}(n)$. But the number of paths we eliminated is exactly

376: $C_{k-2}(n+1)$, as the figure illustrates. This completes 

377: the proof.

378: \hfill $\blacksquare$

379: \end{itemize}

380: \vskip .5 cm

381: \noindent

382: The probability of absorption for $k=3$

383: \be\label{a}

384: P(x=3)=\sum_{n=0}^{\infty} C_{3}(n)p^{n}(1-p)^{n+3}.

385: \ee

386: Substituting for $C_{3}(n)$ its value from (\ref{phin3}),

387: \[

388: C_{3}(n)=C_{2}(n+1)-C_{1}(n+1), 

389: \]

390: we obtain, after some straightforward computation, 

391: \[

392: P(x=3)=\frac{(1-p)^{3}}{p^{3}}

393: \]

394: for $p \geq 1/2$ and $P(x=3)=1$ for $p \leq 1/2$. 

395: \begin{figure}

396: \begin{center}

397: \resizebox{!}{5.4 truecm}{\includegraphics{pp2.eps}}

398: \vskip .3 cm

399: \parbox{13truecm}{\caption{\baselineskip=16 pt\small \label{fig4}

400: The recurrence relation (\ref{phin3}).}

401: }

402: \end{center}

403: \end{figure}

404: \vskip .3 cm

405: \noindent

406: \section{The general case} 

407: \vskip .3 cm

408: \noindent

409: By induction on $k$ we can now prove, as the three cases above suggest,

410: the following result.

411: \vskip .3 cm

412: \noindent

413: {\theorem The probability $P(x=k)$ of absorption of the particle at the

414: origin,

415: given that it starts at $x=k$, is given by 

416: \[

417: P(x=k)=\left ( P(x=1) \right )^{k}. 

418: \]

419: }

420: \begin{itemize}

421: \item[] {\bf Proof.} The probability of absorption of a particle that

422: starts at $x=k+1$ is equal to the probability that it moves one step to

423: the left to $x=k$ times the probability that it is absorbed at the 

424: origin from there, that is $(1-p)P(x=k)$, plus the probability 

425: that the particle moves one step to the right to $x=k+2$ times the

426: probability that it is absorbed from there, that is $ p P(x=k+2)$.

427: Hence, 

428: \[

429: P(x=k+1)=(1-p)P(x=k)+pP(x=k+2),

430: \]

431: from which we get

432: \be\label{p}

433: P(x=k+2)=\frac{1}{p}P(x=k+1)-\frac{(1-p)}{p}P(x=k),

434: \ee

435: The induction step now follows easily from (\ref{p}).

436: \hfill $\blacksquare$

437: \end{itemize}

438: \vskip .8 cm

439: \noindent

440: {\large \bf Acknowledgments\/}

441: I was supported by FAPESP -- Funda\c c\~ao de Amparo \`a Pesquisa do

442: Estado de S\~ao Paulo -- under grant 01/08485-6. I also thank 

443: Pierluigi Contucci for {\it other} path representations.

444: \vspace{.6cm}

445: \addcontentsline{toc}{section}{References}

446: \begin{thebibliography}{1}

447: %\bibitem{WB} W. Burnside, {\it Theory of Probability}, Dover, NY

448: %(1959) p.33

449: \bibitem{FM} F. Mosteller, {\it Fifty Challenging Problems in

450: Probability}, Dover, NY (1987) p. 9 

451: \end{thebibliography}

452: 

453: 

454: 

455: \end{document}

456: