1: % FINAL VERSION AFTER CORRECTIONS IN PROOF 2/5/05

2: \documentclass[12pt]{article}

3: \usepackage{graphicx}

4: \setlength{\textwidth}{16cm}

5: %\setlength{\textwidth}{17cm}

6: %\setlength{\textheight}{24cm}

7: \setlength{\textheight}{22cm}

8: \setlength{\topmargin}{0cm}

9: \setlength{\oddsidemargin}{0cm}

10: \setlength{\parindent}{0cm}

11: 

12: \def\ffrac#1#2{\textstyle{#1\over#2}\displaystyle}

13: 

14: 

15: \begin{document}

16: \pagestyle{myheadings}

17: \parindent 0mm

18: \parskip 6pt

19: \markright{SLE for Theoretical Physicists}

20: 

21: \title{SLE for Theoretical Physicists}

22: 

23: \author{John Cardy\\

24: Rudolf Peierls Centre for Theoretical Physics\\

25:          1 Keble Road, Oxford OX1 3NP,

26:          U.K.\footnote{Address for correspondence.}\\

27: and All Souls College, Oxford.}

28: %

29: \date{May 2005}

30: %

31: \maketitle

32: %

33: \begin{abstract}

34: This article provides an introduction to

35: Schramm(stochastic)-Loewner evolution (SLE) and to its connection

36: with conformal field theory, from the point of view of its

37: application to two-dimensional critical behaviour. The emphasis is

38: on the conceptual ideas rather than rigorous proofs.

39: \end{abstract}

40: %

41: \newpage

42: %

43: 

44: \section{Introduction}

45: %

46: \subsection{Historical overview}

47: %

48: The study of critical phenomena has long been a breeding ground

49: for new ideas in theoretical physics. Such behaviour is

50: characterised by a diverging correlation length and cannot easily

51: be approximated by considering small systems with only a few

52: degrees of freedom. Initially, it appeared that the analytic study

53: of such problems was a hopeless task, although self-consistent

54: approaches such as mean field theory were often successful in

55: providing a semi-quantitative description.

56: 

57: Following Onsager's calculation of the free energy of the square

58: lattice Ising model in 1944, steady progress was made in the exact

59: solution of an ever-increasing number of lattice models in two

60: dimensions. While many of these are physically relevant, and the

61: techniques used have spawned numerous spin-offs in the theory of

62: integrable systems, it is fair to say that these methods have not

63: cast much light on the general nature of the critical state. In

64: addition, virtually no progress has been made from this direction

65: in finding analytic solutions to such simple and important

66: problems as percolation.

67: 

68: An important breakthrough occurred in the late 1960's, with the

69: development of renormalisation group (RG) ideas by Wilson and

70: others. The fundamental realisation was that, in the scaling limit

71: where both the correlation length and all other macroscopic length

72: scales are much larger than that of the microscopic interactions,

73: classical critical systems are equivalent to renormalisable

74: quantum field theories in euclidean space-time. Since there is

75: often only a finite or a denumerable set of such field theories

76: with given symmetries, at a stroke this explained the observed

77: phenomenon of universality: systems with very different

78: constituents and microscopic interactions nevertheless exhibit the

79: same critical behaviour in the scaling limit. This single idea has

80: led to a remarkable unification of the theoretical bases of

81: particle physics, statistical mechanics and condensed matter

82: theory, and has led to extensive cross-fertilisation between these

83: disciplines. These days, a typical paper using the ideas and

84: methods of quantum field theory is as likely to appear in a

85: condensed matter physics journal as in a particle physics

86: publication (although there seems to be a considerable degree of

87: conservatism among the writers of field theory text books in

88: recognising this fact.)

89: 

90: Two important examples of this interdisciplinary flow were the

91: development of lattice gauge theories in particle physics, and the

92: application of conformal field theory (CFT), first developed as a

93: tool in string theory, to statistical mechanics and condensed

94: matter physics. As will be explained later, in two-dimensional

95: classical systems and quantum systems in 1+1 dimensions conformal

96: symmetry is extremely powerful, and has led to a cornucopia of new

97: exact results. Essentially, the RG programme of classifying all

98: suitable renormalisable quantum field theories in two dimensions

99: has been carried through to its conclusion in many cases,

100: providing exact expressions for critical exponents, correlation

101: functions, and other universal quantities. However the

102: geometrical, as opposed to the algebraic, aspects of conformal

103: symmetry are not apparent in this approach.

104: 

105: One minor but nevertheless theoretically influential prediction of

106: these methods was the conjectured crossing formula

107: \cite{JCcrossing} for the probability that, in critical

108: percolation, a cluster should exist which spans between two

109: disjoint segments of the boundary of some simply connected region

110: (a more detailed account of this problem will be given later.)

111: With this result, the simmering unease that mathematicians felt

112: about these methods came to the surface (see, for example, the

113: comments in \cite{Lang}.) What exactly are these renormalised

114: local operators whose correlation functions the field theorists so

115: happily manipulate, according to rules that sometimes seem to be a

116: matter of cultural convention rather than any rigorous logic? What

117: does conformal symmetry really mean? Exactly which object is

118: conformally invariant? And so on. Aside from these deep concerns,

119: there was perhaps also the territorial feeling that percolation

120: theory, in particular, is a branch of probability theory, and

121: should be understood from that point of view, not merely as a

122: by-product of quantum field theory.

123: 

124: Thus it was that a number of pure mathematicians, versed in the

125: methods of probability theory, stochastic analysis and conformal

126: mapping theory, attacked this problem. Instead of trying make

127: rigorous the notions of field theory about local operators, they

128: focused on the random curves which form the boundaries of clusters

129: on the lattice, and on what should be the properties of the

130: measure on such curves in the continuum limit as the lattice

131: spacing approaches zero. The idea of thinking about lattice models

132: this way was not new: in particular in the 1980s it led to the

133: very successful but non-rigorous Coulomb gas approach

134: \cite{Nienhuis} to two-dimensional critical behaviour, whose

135: results parallel and complement those of CFT. However, the new

136: approach focused on the properties of a single such curve,

137: conditioned to start at the boundary of the domain, in the

138: background of all the others. This leads to a very specific and

139: physically clear notion of conformal invariance. Moreover, it was

140: shown by Loewner\cite{Loewner} in the 1920s that any such curve in

141: the plane which does not cross itself can be described by a

142: dynamical process called Loewner evolution, in which the curve is

143: imagined to be grown in a continuous fashion. Instead of

144: describing this process directly, Loewner considered the evolution

145: of the analytic function which conformally maps the region outside

146: the curve into a standard domain. This evolution, and therefore

147: the curve itself, turns out to be completely determined by a real

148: continuous function $a_t$. For random curves, $a_t$ itself is

149: random. (The notation $a_t$ is used rather than $a(t)$ to conform

150: to standard usage in the case when it is a stochastic variable.)

151: Schramm\cite{Schramm} argued that, if the measure on the curve is

152: to be conformally invariant in the precise sense referred to

153: above, the only possibility is that $a_t$ be one-dimensional

154: Brownian motion, with only a single parameter left undetermined,

155: namely the diffusion constant $\kappa$. This leads to stochastic-,

156: or Schramm-, Loewner evolution (SLE). (In the original papers by

157: Schramm et al. the term `stochastic' was used. However in the

158: subsequent literature the `S' has often been taken to stand for

159: Schramm in recognition of his contribution.) It should apply to

160: any critical statistical mechanics model in which it is possible

161: to identify these non-crossing paths on the lattice, as long as

162: their continuum limits obey the underlying conformal invariance

163: property. For only a few cases, including percolation, has it been

164: proved that this property holds, but it is believed to be true for

165: suitably defined curves in a whole class of systems known as

166: O$(n)$ models. Special cases, apart from percolation, include the

167: Ising model, Potts models, the XY model, and self-avoiding walks.

168: They each correspond to a particular choice of $\kappa$.

169: 

170: Starting from the assumption that SLE describes such a single

171: curve in one of these systems, many properties, such as the values

172: of many of the critical exponents, as well as the crossing formula

173: mentioned above, have been rigorously derived in a brilliant

174: series of papers by Lawler, Schramm and Werner (LSW)\cite{LSW}.

175: Together with Smirnov's proof \cite{Smirnov} of the conformal

176: invariance property for the continuum limit of site percolation on

177: the triangular lattice, they give a \em rigorous \em derivation

178: \cite{SmirnovWerner} of the values of the critical exponents for

179: two-dimensional percolation. This represents a paradigm shift in

180: rigorous statistical mechanics, in that results are now being

181: derived directly in the continuum for models for which the

182: traditional lattice methods have, so far, failed.

183: 

184: However, from the point of view of theoretical physics, these

185: advances are  perhaps not so important for being rigorous, as for

186: the new light they throw on the nature of the critical state, and

187: on conformal field theory. In the CFT of the O$(n)$ model, the

188: point where a random curve hits the boundary corresponds to the

189: insertion of a local operator which has a particularly simple

190: property: its correlation functions satisfy linear second-order

191: differential equations\cite{JCsurf}. These equations turn out to

192: be directly related to the Fokker-Planck type equations one gets

193: from the Brownian process which drives SLE. Thus there is a close

194: connection, at least at an operational level, between CFT and SLE.

195: This has been made explicit in a series of papers by Bauer and

196: Bernard\cite{BB} (see also \cite{FriedrichWerner}.) Other

197: fundamental concepts of CFT, such as the central charge $c$, have

198: their equivalence in SLE. This is a rapidly advancing subject, and

199: some of the more recent directions will be mentioned in the

200: concluding section of this article.

201: %

202: \subsection{Aims of this article}

203: The original papers on SLE are mostly both long and difficult,

204: using, moreover, concepts and methods foreign to most theoretical

205: physicists. There are reviews, in particular those by

206: Werner\cite{Wreview} and by Lawler\cite{Lreview} which cover much

207: of the important material in the original papers. These are

208: however written for mathematicians. A more recent review by Kager

209: and Nienhuis\cite{KNreview} describes some of the mathematics in

210: those papers in way more accessible to theoretical physicists, and

211: should be essential reading for any reader who wants then to

212: tackle the mathematical literature. A complete bibliography up to

213: 2003 appears in \cite{KadGruz}.

214: 

215: However, the aims of the present article are more modest. First,

216: it does not claim to be a thorough review, but rather a

217: semi-pedagogical introduction. In fact some of the material,

218: presenting some of the existing results from a slightly different,

219: and hopefully clearer, point of view, has not appeared before in

220: print. The article is directed at the theoretical physicist

221: familiar with the basic concepts of quantum field theory and

222: critical behaviour at the level of a standard graduate textbook,

223: and with a theoretical physicist's knowledge of conformal mappings

224: and stochastic processes. It is not the purpose to prove anything,

225: but rather to describe the concepts and methods of SLE, to relate

226: them to other ideas in theoretical physics, in particular CFT, and

227: to illustrate them with a few simple computations, which, however,

228: will be presented in a thoroughly non-rigorous manner. Thus, this

229: review is most definitely not for mathematicians interested in

230: learning about SLE, who will no doubt cringe at the lack of

231: preciseness in some of the arguments and perhaps be puzzled by the

232: particular choice of material. The notation used will be that of

233: theoretical physics, for example $\langle\cdots\rangle$ for

234: expectation value, and so will the terminology. The word

235: `martingale' has just made its only appearance. Perhaps the

236: largest omission is any account of the central arguments of

237: LSW\cite{LSW} which relate SLE to various aspects of Brownian

238: motion and thus allow for the direct computation of many critical

239: exponents. These methods are in fact related to two-dimensional

240: quantum gravity, whose role in this is already the subject of a

241: recent long article by Duplantier\cite{Dupreview}.

242: 

243: \subsubsection*{Acknowledgments.}

244: One may go only so far in learning a subject like this by reading

245: the original literature, and much of my knowledge, such as it is,

246: comes from seminars by, and informal discussions with, a large

247: number of people. To them I give thanks, as well as apologies if I

248: have occasionally misrepresented or oversimplified their ideas:

249: M.~Aizenman, M.~Bauer, R.~Bauer, V.~Beffara, J.~Dub\'edat,

250: D.~Bernard, M.~den Nijs, B.~Doyon, B.~Duplantier, R.~Friedrich,

251: I.~Gruzberg, M.~Hastings, J.~Jacobsen, L.~Kadanoff, W.~Kager,

252: R.~Kenyon, H.~Kesten, P.~Kleban, J.~Kondev, R.~Langlands,

253: G.~Lawler, B.~Nienhuis, S.~Rohde, Y.~Saint-Aubin, H.~Saleur,

254: O.~Schramm, S.~Sheffield, S.~Smirnov, W.~Werner, D.~Wilson. I also

255: thank B.~Doyon and V.~Riva for comments on the manuscript. This

256: work was begun while the author was a Member of the Institute for

257: Advanced Study, supported by the Ellentuck Fund.

258: %

259: \newpage

260: %

261: \section{Random curves and lattice models.}

262: \label{sec:lattice}

263: \subsection{The Ising and percolation models}

264: \label{sec:ising} In this section, we introduce the lattice models

265: which can be interpreted in terms of random non-intersecting paths

266: on the lattice whose continuum limit will be described by SLE.

267: 

268: The prototype is the Ising model. It is most easily realised on a

269: honeycomb lattice (see Fig.~\ref{fig:hex}).

270: %

271: \begin{figure}[h]

272: \centering

273: \includegraphics[width=8cm]{hex.eps}

274: \caption{\label{fig:hex}\small

275: Ising model on the honeycomb lattice, with loops corresponding to a

276: term in the expansion of (\ref{ZI}). Alternatively, these may be

277: thought of as domain walls of an Ising model on the dual triangular

278: lattice.}

279: \end{figure}

280: %

281: At each site $r$ is an Ising `spin' $s(r)$ which takes the

282: values $\pm1$. The partition function is

283: \begin{equation}

284: \label{ZI} Z_{\rm Ising}={\rm Tr}\,e^{\beta J\sum_{rr'}s(r)s(r')}

285: \propto{\rm Tr}\,\prod_{rr'}\left(1+xs(r)s(r')\right)\,,

286: \end{equation}

287: where $x=\tanh\beta J$, and the sum and product are over all edges

288: joining nearest neighbour pairs of sites. The trace operation is

289: defined as ${\rm Tr}=\prod_r\big(\frac12\sum_{s(r)}\big)$, so that

290: ${\rm Tr}\,s(r)^n=1$ if $n$ is even, and $0$ if it is odd.

291: 

292: At high temperatures ($\beta J\ll 1$) the spins are disordered,

293: and their correlations decay exponentially fast, while at low

294: temperatures ($\beta J\gg1$) there is long-range order: if the

295: spins on the boundary are fixed say, to the value $+1$, then

296: $\langle s(r)\rangle\not=0$ even in the infinite volume limit. In

297: between, there is a critical point. The conventional approach to

298: the Ising model focuses on the behaviour of the correlation

299: functions of the spins. In the scaling limit, they become local

300: operators in a quantum field theory (QFT). Their correlations are

301: power-law behaved at the critical point, which corresponds to a

302: massless QFT, that is a conformal field theory (CFT). From this

303: point of view (as well as exact lattice calculations) it is found

304: that correlation functions like $\langle s(r_1)s(r_2)\rangle$

305: decay at large separations according to power laws

306: $|r_1-r_2|^{-2x}$: one of the aims of the theory is to obtain the

307: values of the exponents $x$ as well as to compute, for example,

308: correlators depending on more than two points.

309: 

310: However, there is an alternative way of thinking about the

311: partition function (\ref{ZI}), as follows: imagine expanding out

312: the product to obtain $2^N$ terms, where $N$ is the total number

313: of edges. Each term may be represented by a subset of edges, or

314: graph $\cal G$, on the lattice, in which, if the term $xs(r)s(r')$

315: is chosen, the corresponding edge $(rr')$ is included in $\cal G$,

316: otherwise it is not. Each site $r$ has either 0, 1, 2 or 3 edges

317: in $\cal G$. The trace over $s(r)$ gives 1 if this number is even,

318: and 0 if it is odd. Each surviving graph is then the union of

319: non-intersecting closed loops (see Fig.~\ref{fig:hex}). In

320: addition, there can be open paths beginning and ending at a

321: boundary. For the time being, we suppress these by imposing `free'

322: boundary conditions, summing over the spins on the boundary. The

323: partition function is then

324: \begin{equation}

325: Z_{\rm Ising}=\sum_{\cal G}\,x^{\rm length}\,,

326: \end{equation}

327: where the length is the total of all the loops in $\cal G$.

328: When $x$ is small, the mean length of a single loop is small. The

329: critical point $x_c$ is signalled by a divergence of this quantity.

330: The low-temperature phase corresponds to $x>x_c$. While in this phase

331: the Ising spins are ordered, and their connected correlation functions\

332: decay exponentially, the loop gas is in fact still critical, in that, for

333: example, the probability that two points lie on the same loop has

334: a power-law dependence on their separation. This is the

335: \em dense \em phase.

336: 

337: The loops in $\cal G$ may be viewed in another way: as \em domain

338: walls \em for another Ising model on the dual lattice, which is a

339: triangular lattice whose sites $R$ lie at the centres of the

340: hexagons of the honeycomb lattice (see Fig.~\ref{fig:hex}). If the

341: corresponding interaction strength of this dual Ising model is

342: $(\beta J)^*$, then the Boltzmann weight for creating a segment of

343: domain wall is $e^{-2(\beta J)^*}$. This should be equated to

344: $x=\tanh(\beta J)$ above. Thus we see that the high-temperature

345: regime of the dual model corresponds to low temperature in the

346: original model, and vice versa. Infinite temperature in the dual

347: model ($(\beta J)^*=0$) means that the dual Ising spins are

348: independent random variables. If we colour each dual site with

349: $s(R)=+1$ black, and white if $s(R)=-1$, we have the problem of

350: \em site percolation \em on the triangular lattice, critical

351: because $p_c=\frac12$ for that problem. Thus the curves with $x=1$

352: correspond to percolation cluster boundaries. (In fact in the

353: scaling limit this is believed to be true throughout the dense

354: phase $x>x_c$.)

355: 

356: So far we have discussed only closed loops. Consider the spin-spin

357: correlation function

358: \begin{equation}

359: \langle s(r_1)s(r_2)\rangle= {{\rm

360: Tr}\,s(r_1)s(r_2)\prod_{rr'}\left(1+xs(r)s(r')\right) \over{\rm

361: Tr}\,\prod_{rr'}\left(1+xs(r)s(r')\right)}\,,

362: \end{equation}

363: where the sites $r_1$ and $r_2$ lie on the boundary. Expanding out

364: as before, we see that the surviving graphs in the numerator each

365: have a single edge coming into $r_1$ and $r_2$. There is therefore

366: a single open path $\gamma$ connecting these points on the

367: boundary (which does not intersect itself nor any of the closed

368: loops.) In terms of the dual variables, such a single open curve

369: may be realised by specifying the spins $s(R)$ on all the dual

370: sites on the boundary to be $+1$ on the part of the boundary

371: between $r_1$ and $r_2$ (going clockwise) and $-1$ on the

372: remainder. There is then a single domain wall connecting $r_1$ to

373: $r_2$. {\it SLE describes the continuum limit of such a curve

374: $\gamma$.}

375: 

376: Note that we could also choose $r_2$ to lie in the interior. The

377: continuum limit of such curves is then described by radial SLE

378: (Sec.~\ref{sec:radialsle}).

379: 

380: \subsubsection{Exploration process} \label{sec:exploration}

381: An important property of the ensemble of curves $\gamma$ on the

382: lattice is that, instead of generating a configuration of all the

383: $s(R)$ and then identifying the curve, it may be constructed

384: step-by-step as follows (see Fig.~\ref{fig:exploration}).

385: %

386: \begin{figure}

387: \centering

388: \includegraphics[width=8cm]{explore.eps}

389: \caption{\label{fig:exploration}\small

390: The exploration process for the Ising model. At each step the walk

391: turns L or R according to the value of the spin in front of it. The

392: relative probabilities are determined by the expectation value of this

393: spin given the fixed spins either side of the walk up to this time. The

394: walk never crosses itself and never gets trapped.}

395: \end{figure}

396: %

397: Starting

398: from $r_1$, at the next step it should turn R or L according to

399: whether the spin in front of it is $+1$ or $-1$. For independent

400: percolation, the probability of either event is $\frac12$, but for

401: $x<1$ it depends on the values of the spins on the boundary.

402: Proceeding like this, the curve will grow, with all the dual sites

403: on its immediate left taking the value $+1$, and those its right

404: the value $-1$. The relative probabilities of the path turning R

405: or L at a given step depend on the expectation value of the spin

406: on the site $R$ immediately in front of it, given the values of

407: the spins already determined, that is, given the path up to that

408: point. Thus the relative probabilities that the path turns R or L

409: are completely determined by the domain and the path up to that

410: point. This implies the crucial

411: 

412: {\bf Property 2.1} (lattice version). Let $\gamma_1$ be the part

413: of the total path covered after a certain number of steps. Then

414: the conditional probability distribution of the remaining part of

415: the curve, given $\gamma_1$, is the same as the unconditional

416: distribution of a whole curve, starting at the tip and ending at

417: $r_2$ in the domain ${\cal D}\setminus\gamma_1$.

418: 

419: In the Ising model, for example, if we already know part of the domain wall,

420: the rest of it can be considered as a complete domain

421: wall in a new region in which the left and right sides of the existing part

422: form part of the boundary.

423: This means the path is a history-dependent random walk. It can

424: be seen (Fig.~\ref{fig:exploration}) that when the growing tip

425: $\tau$

426: approaches an earlier section of the path, it must always turn away

427: from it: the tip never gets trapped.

428: There is always at least one path on the lattice from

429: the tip $\tau$ to the final point $r_2$.

430: 

431: \subsection{O$(n)$ model}\label{sec:Onmodel}

432: The loop gas picture of the Ising and percolation models may

433: simply be generalised by counting each closed loop with a fugacity

434: $n$:

435: \begin{equation}\label{On1}

436: Z_{{\rm O}(n)}=\sum_{\cal G}\,x^{\rm length}\,n^{\rm number\ of\

437: loops}\,.

438: \end{equation}

439: This is called the O$(n)$ model, for the reason that it gives the

440: partition function for $n$-component spins

441: ${\bf s}(r)=(s_1(r),\ldots s_n(r))$

442: with

443: \begin{equation}\label{On2}

444: Z_{{\rm O}(n)}={\rm Tr}\,\prod_{rr'}\left(1+ x{\bf s}(r)\cdot{\bf

445: s}(r')\right)\,,

446: \end{equation}

447: where ${\rm Tr}\,s_a(r)s_b(r)=\delta_{ab}$. Following the same

448: procedure as before we obtain the same set of closed loops (and

449: open paths) except that, on summing over the last spin in each

450: closed loop, we get a factor $n$. The model is called O$(n)$

451: because of its symmetry under rotations of the spins. The version

452: (\ref{On2}) makes sense only when $n$ is a positive integer (and

453: note that the form of the partition function is different from

454: that of the conventional O$(n)$ model, where the second term is

455: exponentiated.) The form in (\ref{On1}) is valid for general

456: values of $n$, and it gives a probability measure on the loop gas

457: for real $n\geq0$. However, the dual picture is useful only for

458: $n=1$ and $n=2$ (see below.) As for the case $n=1$, there is a

459: critical value $x_c(n)$ at which the mean loop length diverges.

460: Beyond this, there is a dense phase.

461: 

462: Apart from $n=1$, other important physical values of $n$ are:

463: \begin{itemize}

464: \item $n=2$. In this case we can view each loop as being oriented

465: in either a clockwise or anti-clockwise sense, giving it an overall

466: weight 2. Each loop configuration then corresponds to a

467: configuration of integer valued\ \em height \em variables $h(R)$

468: on the dual lattice, with the convention that the nearest

469: neighbour difference $h(R')-h(R)$ takes the values 0, $+1$ or $-1$

470: according to whether the edge crossed by $RR'$ is unoccupied,

471: occupied by an edge oriented at $90^{\circ}$ to $RR'$, or at

472: $-90^{\circ}$. (That is, the current running around each loop is

473: the lattice curl of $h$.) The variables $h(R)$ may be pictured as

474: the local height of a crystal surface. In the low-temperature

475: phase (small $x$) the surface is smooth: fluctuations of the

476: height differences decay exponentially with separation. In the

477: high-temperature phase it is rough: they grow logarithmically. In

478: between is a roughening transition. It is believed that relaxing

479: the above restriction on the height difference $h(R')-h(R)$ does

480: not change the universality class, as long as large values of this

481: difference are suppressed, for example using the weighting

482: $\exp\big[-\beta\big(h(R')-h(R)\big)^2\big]$. This is the discrete

483: Gaussian model. It is dual to a model of 2-component spins with

484: O$(2)$ symmetry called the XY model. \item $n=0$. In this case,

485: closed loops are completely suppressed, and we have a single

486: non-self-intersecting path connecting $r_1$ and $r_2$, weighted by

487: its length. Thus, all paths of the same length are counted with

488: equal weight. This is the self-avoiding walk problem, which is

489: supposed to describe the behaviour of long flexible polymer

490: chains. As $x\to x_c-$, the mean length diverges. The region

491: $x>x_c$ is the dense phase, corresponding to a long polymer whose

492: length is of the order of the area of the box, so that it has

493: finite density. \item $n=-2$ corresponds to the loop-erased random

494: walk. This is an ordinary random walk in which every loop, once it

495: is formed, is erased. Taking $n=-2$ in the $O(n)$ model of

496: non-intersecting loops has this effect.

497: \end{itemize}

498: %

499: \subsection{Potts model.}

500: Another important model which may described in terms of random

501: curves in the $Q$-state Potts model. This is most easily considered

502: on square lattice, at each site of which is a variable $s(r)$ which can take

503: $Q$ (initially a positive integer) different values. The partition function

504: is

505: \begin{equation}

506: Z_{\rm Potts}={\rm Tr}\,e^{\beta J\sum_{rr'}\delta_{s(r),s(r')}}

507: \propto{\rm

508: Tr}\,\prod_{rr'}\left(1-p+p\delta_{s(r),s(r')}\right)\,,

509: \end{equation}

510: with $e^{\beta J}=(1-p)^{-1}$. The product may be expanded in a

511: similar way to the case of the Ising model. All possible graphs

512: $\cal G$ will appear. Within each connected component of $\cal G$

513: the Potts spins must be equal, giving rise to a factor $Q$ when

514: the trace is performed. The result is

515: \begin{equation}

516: Z_{\rm Potts}=\sum_{\cal G}\,p^{|{\cal G}|}(1-p)^{|\overline{\cal

517: G}|} Q^{||{\cal G}||}\,,

518: \end{equation}

519: where $|{\cal G}|$ is the number of edges in $\cal G$,

520: $|\overline{\cal G}|$ the number in its complement, and $||{\cal

521: G}||$ is the number of connected components of $\cal G$, which are

522: called Fortuin-Kasteleyn (FK) clusters. This is the \em random

523: cluster \em representation of the Potts model. When $p$ is small,

524: the mean cluster size is small. As $p\to p_c$, it diverges, and

525: for $p>p_c$ there is an infinite cluster which contains a finite

526: fraction of all the sites in the lattice. It should be noted that

527: these FK clusters are \em not \em the same as the spin clusters

528: within which the original Potts spins all take the same value.

529: 

530: The limit $Q\to1$ gives another realisation of percolation -- this

531: time bond percolation on the square lattice. For $Q\to0$ there is

532: only one cluster. If at the same time $x\to0$ suitably, all loops

533: are suppressed and the only graphs $\cal G$ which contribute are

534: \em spanning trees\em, which contain every site of the lattice. In

535: the Potts partition function each possible spanning tree is

536: counted with the same weight, corresponding to the problem of

537: uniform spanning trees (UST). The ensemble of paths on USTs

538: connecting two points $r_1$ and $r_2$ turns out to be be that of

539: loop-erased random walks.

540: %

541: \begin{figure}

542: \centering

543: \includegraphics[width=8cm]{medial.eps}

544: \caption{\label{fig:medial}\small Example of FK clusters (heavy

545: lines) in the random cluster representation of the Potts model,

546: and the corresponding set of dense loops (medium heavy) on the

547: medial lattice. The loops never cross the edges connecting sites

548: in the same cluster.}

549: \end{figure}

550: %

551: 

552: The random cluster model may be realised as a gas of dense loops

553: in the way illustrated in Fig.~\ref{fig:medial}. These loops lie

554: on the medial lattice, which is also square but has twice the

555: number of sites. It may be shown that, at $p_c$, the weights for

556: the clusters are equivalent to counting each loop with a fugacity

557: $\sqrt Q$. Thus the boundaries of the critical FK clusters in the

558: $Q$-state Potts model are the same in the scaling limit (if it

559: exists) as the closed loops of the dense phase of the O$(n)$

560: model, with $n=\sqrt Q$.

561: 

562: To generate an open path in the random cluster model

563: connecting sites $r_1$ and $r_2$ on the boundary we must choose

564: `wired' boundary conditions, in which $p

565: =1$ on all the edges

566: parallel to the boundary, from $r_1$ to $r_2$, and free boundary

567: conditions, with $p=0$, along the remainder.

568: 

569: \subsection{Coulomb gas methods}\label{sec:coulomb}

570: Many important results concerning the O$(n)$ model can be derived

571: in a non-rigorous fashion using so-called Coulomb gas methods. For

572: the purposes of comparison with later results from SLE, we now

573: summarise these methods and collect a few relevant formulae. A

574: much more complete discussion may be found in the review by

575: Nienhuis\cite{Nienhuis}.

576: 

577: We assume that the boundary conditions on the O$(n)$ spins are free,

578: so that the partition function is a sum over closed loops only.

579: First orient each loop at random. Rather than giving clockwise and

580: anti-clockwise orientations the same weight $n/2$, give them complex

581: weights $e^{\pm 6i\chi}$, where $n=e^{6i\chi}+e^{-6i\chi}

582: =2\cos6\chi$. These may be taken into account, on the honeycomb lattice,

583: by assigning a weight $e^{\pm i\chi}$ at each vertex where an oriented

584: loop turns R (resp.~L). This transforms the non-local factors of $n$

585: into local (albeit complex) weights depending only on the local

586: configuration at each vertex.

587: 

588: Next transform to the height variables described above. By convention,

589: the heights are taken to be integer multiples of $\pi$. The local weights

590: at each vertex now depend only on the differences of the three adjacent

591: heights. The crucial assumption of the Coulomb gas approach is that,

592: under the RG, this model flows to one in which the lattice can be replaced

593: by a continuum, and the heights go over into a gaussian free field, with

594: partition function $Z=\int e^{-S[h]}[dh]$, where

595: \begin{equation}\label{gff}

596: S=(g/4\pi)\int(\nabla h)^2d^2r\,.

597: \end{equation}

598: As it stands, this is a simple free field theory.

599: The height fluctuations grow logarithmically:

600: $\langle\big(h(r_1)-h(r_2)\big)^2\rangle\sim(2/g)\ln|r_1-r_2|$,

601: and the correlators of exponentials of the height decay with power

602: laws:

603: \begin{equation}

604: \langle e^{iqh(r_1)}\,e^{-iqh(r_2)}\rangle\sim

605: |r_1-r_2|^{-2x_q}\,,

606: \end{equation}

607: where $x_q=q^2/2g$. All the subtleties come from the combined

608: effects of the phase factors and the boundaries or the topology.

609: This is particularly easy to see if we consider the model on a

610: cylinder of circumference $\ell$ and length $L\gg\ell$. In the

611: simple gaussian model (\ref{gff}) the correlation function between

612: two points a distance $L$ apart along the cylinder decays as

613: $\exp\big(-2\pi x_qL/\ell\big)$. However, if $\chi\not=0$, loops

614: which wrap around the cylinder are not counted correctly by the

615: above prescription, because the total number of left turns minus

616: right turns is then zero. We may arrange the correct factors by

617: inserting $e^{\pm 6i\chi h/\pi}$ at either end of the cylinder.

618: This has the effect of modifying the partition function: one finds

619: $\ln Z\sim (\pi c/6)(L/\ell)$ with

620: \begin{equation}\label{eqn:c6}

621: c=1-6{(6\chi/\pi)^2\over g}\,.

622: \end{equation}

623: This dependence of the partition function is one way of

624: determining the so-called central charge of the corresponding CFT

625: (Sec.~\ref{sec:cft}). The charges at each end of the cylinder also

626: modify the scaling dimension $x_q$ to

627: $(1/2g)\big((q-6i\chi/\pi)^2- (6i\chi/\pi)^2\big)$.

628: 

629: The value of $g$ may be fixed \cite{Kondev} in terms of the

630: original discreteness of the height variables as follows: adding a

631: term $-\lambda\int\cos2h\,d^2r$ to $S$ in (\ref{gff}) ensures

632: that, in the limit $\lambda\to\infty$, $h$ will be an integer

633: multiple of $\pi$. For this deformation not to affect the critical

634: behaviour, it must be marginal in the RG sense, which means that

635: it must have scaling dimension $x_2=2$. This condition then

636: determines $g=1-6\chi/\pi$.

637: 

638: \subsubsection{Winding angle distribution}\label{sec:winding1}

639: A simple property which can be inferred from the Coulomb gas

640: formulation is the winding angle distribution. Consider a cylinder

641: of circumference $2\pi$ and a path that winds around it. What the

642: probability that it winds through an angle $\theta$ around the

643: cylinder while it moves a distance $L\gg1$ along the axis? This

644: will correspond to a height difference $\Delta h=\pi(\theta/2\pi)$

645: between the ends of the cylinder, and therefore an additional free

646: energy $(g/4\pi)(2\pi L)(\theta/2L)^2$. The probability density is

647: therefore

648: \begin{equation}

649: P(\theta)\propto \exp(-g\theta^2/8L)\,,

650: \end{equation}

651: so that $\theta$ is normally distributed with variance $(4/g)L$.

652: This result will be useful later (Sec.~\ref{sec:radialsle}) for

653: comparison with SLE.

654: 

655: \subsubsection{$N$-leg exponent}\label{Nleg1}

656: As a final simple exponent prediction, consider the correlation

657: function $\langle\Phi_N(r_1)\Phi_N(r_2)\rangle$ of the $N$-leg

658: operator, which in the language of the O$(n)$ model is

659: $\Phi_N=s_{a_1}\cdots s_{a_N}$, where none of the indices are

660: equal. It gives the probability that $N$ mutually non-intersecting

661: curves connect the two points. Taking them a distance $L\gg\ell$

662: apart along the cylinder, we can choose to orient them all in the

663: same sense, corresponding to a discontinuity in $h$ of $N\pi$ in

664: going around the cylinder. Thus we can write $h=\pi Nv/\ell+\tilde

665: h$, where $0\leq v<\ell$ is the coordinate around the cylinder,

666: and $\tilde h(v+\ell)=\tilde h(v)$. This gives

667: \begin{equation}\label{eqn:Nleg}

668: \langle\Phi_N(r_1)\Phi_N(r_2)\rangle\sim

669: e^{-(g/4\pi)(N\pi/\ell)^2L+(\pi L/6\ell)-(\pi cL/6\ell)}\,.

670: \end{equation}

671: The second term in the exponent comes from the integral over the

672: fluctuations $\tilde h$, and the last from the partition function.

673: They differ because in the numerator, once there are curves

674: spanning the length of the cylinder, loops around it, which give

675: the correction term in (\ref{eqn:c6}), are forbidden.

676: (\ref{eqn:Nleg}) then gives

677: \begin{equation}

678: x_N=(gN^2/8)-(g-1)^2/2g\,.

679: \end{equation}

680: %

681: \newpage

682: %

683: \section{SLE}\label{sec:sle}

684: \subsection{The postulates of SLE.}\label{sec:postulates}

685: SLE gives a description of the continuum limit of the lattice

686: curves connecting two points on the boundary of a domain $\cal D$

687: which were introduced in Sec.~\ref{sec:lattice}. The idea is to

688: define a \em measure \em $\mu(\gamma;{\cal D},r_1,r_2))$ on these

689: continuous curves. (Note that the notion of a probability density

690: of such objects does not make sense, but the more general concept

691: of a measure does.)

692: 

693: There are two basic properties of this

694: continuum limit which must either be assumed, or, better, proven to hold

695: for a particular lattice model.

696: The first is the continuum version of Property 1:

697: 

698: {\bf Property 3.1} (continuum version). Denote the curve by

699: $\gamma$, and divide it into two disjoint parts: $\gamma_1$ from

700: $r_1$ to $\tau$, and $\gamma_2$ from $\tau$ to $r_2$. Then the

701: conditional measure $\mu(\gamma_2|\gamma_1;{\cal D},r_1,r_2)$ is

702: the same as $\mu(\gamma_2;{\cal D}\setminus\gamma_1,\tau,r_2)$.

703: 

704: This property we expect to be true for the scaling limit

705: of all such curves in the

706: O$(n)$ model (at least for $n\geq0$), even away from the critical

707: point. However the second property encodes the notion of conformal

708: invariance, and it should be valid, if at all, only at $x=x_c$

709: and, separately, for $x>x_c$.

710: 

711: {\bf Property 3.2} (conformal invariance.) Let $\Phi$ be a

712: conformal mapping of the interior of the domain $\cal D$ onto the

713: interior of ${\cal D}'$, so that the points $(r_1,r_2)$ on the

714: boundary of $\cal D$ are mapped to points $(r_1',r_2')$ on the

715: boundary of ${\cal D}'$. The measure $\mu$ on curves in $\cal D$

716: induces a measure $\Phi*\mu$ on the image curves in ${\cal D}'$.

717: The conformal invariance property states that this is the same as

718: the measure which would be obtained as the continuum limit of

719: lattice curves from $r_1'$ to $r_2'$ in ${\cal D}'$. That is

720: \begin{equation}

721: (\Phi*\mu)(\gamma;{\cal D},r_1,r_2)= \mu(\Phi(\gamma);{\cal

722: D}',r_1',r_2')\,.

723: \end{equation}

724: %

725: \subsection{Loewner's equation}\label{sec:loewner}

726: We have seen that, on the lattice, the curves $\gamma$ may be

727: `grown' through a discrete exploration process. The Loewner

728: process is the continuum version of this. Because of Property 2 it

729: suffices to describe this in a standard domain $\cal D$, which is

730: taken to be the upper half plane $\bf H$, with the points $r_1$

731: and $r_2$ being the origin and infinity respectively.

732: 

733: The first thing to notice is that, although on the honeycomb

734: lattice the growing path does not intersect itself, in the

735: continuum limit it might (although it still should not cross

736: itself.) This means that there may be regions enclosed by the path

737: which are not on the path but nevertheless are not reachable from

738: infinity without crossing it. We call the union of the set of such

739: points, together with the curve itself, up to time $t$, the \em

740: hull \em $K_t$. (This is a slightly different usage of this term

741: from that in the physics percolation literature.) It is the

742: complement of the connected component of the half plane which

743: includes $\infty$, itself denoted by ${\bf H}\setminus K_t$. See

744: Fig.~\ref{fig:hull}.

745: %

746: \begin{figure}

747: \centering

748: \includegraphics[width=6cm]{hull.eps}

749: \caption{\label{fig:hull}\small Schematic view of a trace and its

750: hull.}

751: \end{figure}

752: %

753: 

754: Another property which often holds in the half-plane is that of

755: \em reflection invariance\em: the distribution of lattice paths

756: starting from the origin and ending at $\infty$ is invariant under

757: $x\to-x$. For the lattice paths in the $O(n)$ model discussed in

758: Sec.~\ref{sec:Onmodel} this follows from the symmetry of the

759: underlying weights, but for the boundaries of the FK clusters in

760: the Potts model it is a consequence of duality. Not all simple

761: curves in lattice models have this property. For example if we

762: consider the 3-state Potts model in which the spins on the

763: negative and positive real axes are fixed to different values,

764: there is a simple lattice curve which forms the outer boundary of

765: the spin cluster containing the positive real axis. This is not

766: the same as the boundary of the spin cluster containing the

767: negative real axis, and it is not in general symmetric under

768: reflections.

769: 

770: Since ${\bf H}\setminus K_t$ is simply connected, by the Riemann

771: mapping theorem it can be mapped into the standard domain $\bf H$

772: by an analytic function $g_t(z)$. Because this preserves the real

773: axis outside $K_t$ it is in fact real analytic. It is not unique,

774: but can be made so by imposing the behaviour as $z\to\infty$

775: \begin{equation}\label{eqn:asymp}

776: g_t(z)\sim z+O(1/z)\,.

777: \end{equation}

778: It can be shown that, as the path grows, the coefficient of $1/z$

779: is monotonic increasing (essentially it is the electric dipole

780: moment of $K_t$ and its mirror image in the real axis.) Therefore

781: we may reparametrise time so that this coefficient is $2t$. (The

782: factor 2 is conventional.) Note that the length of the curve is

783: not be a useful parametrisation in the continuum limit, since the

784: curve is a fractal.

785: 

786: The function $g_t(z)$ maps the whole boundary of $K_t$ onto part

787: of the real axis. In particular, it maps the growing tip $\tau_t$

788: to a real point $a_t$. Any point on the real axis outside $K_t$

789: remains on the real axis. As the path grows, the point $a_t$ moves

790: on the real axis. For the path to describe a curve, it must always

791: grow only at its tip, and this means that the function $a_t$ must

792: be continuous, but not necessarily differentiable.

793: 

794: A simple but instructive example is when $\gamma$ is a straight

795: line growing vertically upwards from a fixed point $a$. In this

796: case

797: \begin{equation}\label{eqn:example}

798: g_t(z)=a+\big((z-a)^2+4t\big)^{1/2}\,.

799: \end{equation}

800: This satisfies (\ref{eqn:asymp}), and $\tau_t=2i\sqrt t$. More

801: complicated deterministic examples can be found\cite{KadKager}. In

802: particular, $a_t\propto t^{1/2}$ describes a straight line growing

803: at a fixed angle to the real axis.

804: 

805: Loewner's idea \cite{Loewner} was to describe the path $\gamma$

806: and the evolution of the tip $\tau_t$ in terms of the evolution of

807: the conformal mapping $g_t(z)$. It turns out that the equation of

808: motion for $g_t(z)$ is simple:

809: \begin{equation}\label{eqn:loewner}

810: {dg_t(z)\over dt}={2\over g_t(z)-a_t}\,.

811: \end{equation}

812: This is Loewner's equation. The idea of the proof is

813: straightforward. Imagine evolving the path for a time $t$, and

814: then for a further short time $\delta t$. The image of

815: $K_{t+\delta t}$ under $g_t$ is a short vertical line above the

816: point $a_t$ on the real axis. Thus we can write, using

817: (\ref{eqn:example})

818: \begin{equation}

819: g_{t+\delta t}(z)\approx a_t+\big((g_t(z)-a_t)^2+4\delta

820: t\big)^{1/2}\,.

821: \end{equation}

822: Differentiating with respect to $\delta t$ and then letting

823: $\delta t\to0$, we obtain (\ref{eqn:loewner}).

824: 

825: Note that, even if $a_t$ is not differentiable (as is the case for

826: SLE), (\ref{eqn:loewner}) gives for each point $z_0$ a solution

827: $g_t(z_0)$ which is differentiable with respect to $t$, up to the

828: time when $g_t(z_0)=a_t$. This is the time when $z_0$ is first

829: included in $K_t$. However, it is sometimes (see

830: Sec.~\ref{sec:cft}) useful to normalise the Loewner mapping

831: differently, defining $\hat g_t(z)=g_t(z)-a_t$, which always maps

832: the growing tip $\tau_t$ to the origin. If $a_t$ is not

833: differentiable, neither is $\hat g_t$, and the Loewner equation

834: should be written in differential form as $d\hat g_t=(2dt/\hat

835: g_t)-da_t$.

836: 

837: Given a growing path, we can determine the hull $K_t$ and hence,

838: in principle, the function $g_t(z)$ and thereby $a_t=g_t(\tau_t)$.

839: Conversely, given $a_t$ we can integrate (\ref{eqn:loewner}) to

840: find $g_t(z)$ and hence in determine the curve (although proving

841: that this inverse problem gives a curve is not easy.)

842: 

843: \subsection{Schramm-Loewner Evolution}\label{sec:schramm}

844: In the case that we are interested in, $\gamma$ is a random curve,

845: so that $a_t$ is a random continuous function. What is the measure

846: on $a_t$? This is answered by the following remarkable result, due

847: to Schramm\cite{Schramm}:

848: 

849: {\bf Theorem.} \em If Properties 3.1 and 3.2 hold, together with

850: reflection symmetry, then $a_t$ is proportional to a standard

851: Brownian motion.\em

852: 

853: That is

854: \begin{equation}

855: a_t=\sqrt\kappa B_t\,,

856: \end{equation}

857: so that $\langle a_t\rangle=0$,

858: $\langle\big(a_{t_1}-a_{t_2}\big)^2\rangle=\kappa|t_1-t_2|$. The

859: only undetermined parameter is $\kappa$, the diffusion constant.

860: It will turn out that different values of $\kappa$ correspond to

861: different universality classes of critical behaviour.

862: 

863: The idea behind the proof is once again simple. As before,

864: consider growing the curve for a time $t_1$, giving $\gamma_1$,

865: and denote the remainder $\gamma\setminus\gamma_1=\gamma_2$.

866: Property 3.1 tells us that the conditional measure on $\gamma_2$

867: given $\gamma_1$ is the same as the measure on $\gamma_2$ in the

868: domain ${\bf H}\setminus K_{t_1}$, which, by Property 3.2, induces

869: the same measure on $g_{t_1}(\gamma_2)$ in the domain ${\bf H}$,

870: shifted by $a_{t_1}$ (see Fig.~\ref{fig:loewner}).

871: %

872: \begin{figure}

873: \centering

874: \includegraphics[width=8cm]{loewner.eps}

875: \caption{\label{fig:loewner}\small A hull evolved from $a_0$ for

876: time $t_1$, then to infinity. The measure on the image of the rest

877: of the curve under $g_{t_1}$ is the same, according to the

878: postulates of SLE, as a hull evolved from $a_{t_1}$ to $\infty$.}

879: \end{figure}

880: %

881: In terms of the function $a_t$ this means that the probability law

882: of $a_t-a_{t_1}$, for $t>t_1$, is the same as the law of

883: $a_{t-t_1}$. This implies that all the increments $\Delta_n \equiv

884: a_{(n+1)\delta t}-a_{n\delta t}$ are independent identically

885: distributed random variables, for all $\delta t>0$. The only

886: process that satisfies this is Brownian motion with a possible

887: drift term: $a_t=\sqrt\kappa B_t+\alpha t$. Reflection symmetry

888: then implies that $\alpha=0$.

889: 

890: \subsection{Simple Properties of SLE}\label{sec:simpleproperties}

891: \subsubsection{Phases of SLE}\label{sec:phases}

892: Many of the results discussed in this section have been proved by

893: Rohde and Schramm\cite{RS}. First we address the question of how

894: the trace (the trajectory of $\tau_t$) looks for different values

895: of $\kappa$. For $\kappa=0$, it is a vertical straight line. As

896: $\kappa$ increases, the trace should randomly turn to the L or R

897: more frequently. However, it turns out that there are qualitative

898: differences at critical values of $\kappa$. To see this, let us

899: first study the process on the real axis. Let

900: $x_t=g_t\big(x_0\big)-a_t$ be the distance between the image at

901: time $t$ of a point which starts at $x_0$ and the image $a_t$ of

902: the tip. It obeys the stochastic equation

903: \begin{equation}\label{eqn:bessel}

904: dx_t={2dt\over x_t}-\sqrt\kappa dB_t\,.

905: \end{equation}

906: Physicists often write such an equation as $\dot x=(2/x)-\eta_t$

907: where $\eta_t$ is `white noise' of strength $\kappa$. Of course

908: this does not make sense since $x_t$ is not differentiable. Such

909: equations are always to be interpreted in the `Ito sense', that

910: is, as the limit as $\delta t\to0$ of the forward difference

911: equation $x_{t+\delta t}\approx x_t+(2\delta

912: t/x_t)+\int_t^{t+\delta t}\eta_{t'}dt'$.

913: 

914: (\ref{eqn:bessel}) is known as the Bessel process. (If we set

915: $R_t=(D-1)^{1/2}x_t/2$ and $\kappa^2=4/(D-1)$ it describes the

916: distance $R_t$ from the origin of a Brownian particle in $D$

917: dimensions.) The point $x_t$ is repelled from the origin but it is

918: also subject to a random force. Its ultimate fate can be inferred

919: from the following crude argument: if we ignore the random force,

920: $x_t^2\sim 4t$, while, in the absence of the repulsive term,

921: $\langle x_t^2\rangle\sim\kappa t$. Thus for $\kappa<4$ the

922: repulsive force wins and the particle escapes to infinity, while

923: for $\kappa>4$ the noise dominates and the particle collides with

924: the origin in finite time (at which point the equation breaks

925: down.) A more careful analysis confirms this.

926: %

927: \begin{figure}

928: \centering

929: \includegraphics[width=5cm]{enclose.eps}

930: \caption{\label{fig:enclose}\small The trace is about to hit the

931: axis at $x_0$ and enclose a region. At the time this happens, the

932: whole region including the point $x_0$ is mapped by $g_t$ to the

933: same point $a_t$.}

934: \end{figure}

935: %

936: %

937: What does this collision signify in  terms of the behaviour of the

938: trace? In Fig.~\ref{fig:enclose} we show a trace which is about to

939: hit the real axis at the point $x_0$, thus engulfing a whole

940: region. This is visible from infinity only through a very small

941: opening, which means that, under $g_t$, it gets mapped to a very

942: small region. In fact, as the tip $\tau_t$ approaches $x_0$, the

943: size of the image of this region shrinks to zero. When the gap

944: closes, the whole region enclosed by the trace, as well as

945: $\tau_t$ and $x_0$, are mapped in to the single point $a_t$, which

946: means, in particular, that $x_t\to0$. The above argument shows

947: that for $\kappa<4$ this never happens: the trace never hits the

948: real axis (with probability 1.) For the same reason, it neither

949: hits itself. Thus for $\kappa<4$ the trace $\gamma$ is a \em

950: simple curve\em.

951: 

952: The opposite is true for $\kappa>4$: points on the real axis are

953: continually colliding with the image $a_t$ of the tip. This means

954: that the trace is continually touching both itself and the real

955: axis, at the same time engulfing whole regions. Moreover, since it

956: is self-similar object, it does this on all scales, an infinite

957: number of times within any finite neighbourhood! Eventually the

958: trace swallows the whole half plane: every point is ultimately

959: mapped into $a_t$. For $\kappa<4$ only the points on the trace

960: itself suffer this fate. The case $\kappa=4$ is more tricky: in

961: fact the trace is then also a simple curve.

962: 

963: When $\kappa$ is just above 4, the images of points on the real

964: axis under $g_t$ collide with $a_t$ only when there happen to be

965: rare events when the random force is strong enough to overcome the

966: repulsion. When this happens, whole segments of the real axis are

967: swallowed at one time, corresponding to the picture described

968: above. Conversely, for large $\kappa$, the repulsive force is

969: negligible except for very small $x_t$. In that case, two

970: different starting points move with synchronised Brownian motions

971: until the one which started off closer to the origin is swallowed.

972: Thus the real line is eaten up in a continuous fashion rather than

973: piecemeal. There are no finite regions swallowed by the trace

974: which are not on the trace itself. This means that the trace is

975: \em space-filling\em: $\gamma$ intersects every neighbourhood of

976: every point in the upper half plane. We shall argue later

977: (Sec.~\ref{sec:fractaldim}) that the fractal dimension of the

978: trace is $d_f=1+\kappa/8$ for $\kappa\leq8$ and 2 for

979: $\kappa\geq8$. Thus it becomes space-filling for all

980: $\kappa\geq8$.

981: 

982: \subsubsection{SLE duality}\label{sec:duality}

983: For $\kappa>4$ the curve continually touches itself and therefore

984: its hull $K_t$ contains earlier portions of the trace (see

985: Fig.~\ref{fig:hull}). However, the \em frontier \em of $K_t$ (i.e.

986: the boundary of ${\bf H}\setminus K_t$, minus any portions of the

987: real axis), is by definition a simple curve. A beautiful result,

988: first suggested by Duplantier\cite{Dupduality}, and proved by

989: Beffara\cite{Beffaraduality} for the case $\kappa=6$, is that

990: locally this curve is an SLE$_{\tilde\kappa}$, with

991: \begin{equation}

992: \tilde\kappa=16/\kappa\,.

993: \end{equation}

994: For example, the exterior of a percolation cluster contains many

995: `fjords' which, on the lattice, are connected to the main ocean by

996: a neck of water which is only a finite number of lattice spacings

997: wide. These are sufficiently frequent and the fjords

998: macroscopically large that they survive in the continuum limit.

999: SLE$_{6}$ describes the boundaries of the clusters, including the

1000: coastline of all the fjords. However, the coastline as seen from

1001: the ocean is a simple curve, which is locally SLE$_{8/3}$, the

1002: same as that conjectured for a self-avoiding walk. This suggests,

1003: for example, that locally the frontier of a percolation cluster

1004: and a self-avoiding walk are the same in the scaling limit. In

1005: Sec.~\ref{sec:cft} we show that SLE$_\kappa$ and

1006: SLE$_{\tilde\kappa}$ correspond to CFTs with the same value of the

1007: central charge $c$.

1008: 

1009: 

1010: \subsection{Special values of $\kappa$}

1011: \subsubsection{Locality}\label{sec:locality}

1012: [This subsection and the next are more technical and may be

1013: omitted at a first reading.] We have defined SLE in terms of

1014: curves which connect the origin and infinity in the upper half

1015: plane. Property 2 then allows us to define it for any pair of

1016: boundary points in any simply connected domain, by a conformal

1017: mapping. It is interesting to study how the variation of the

1018: domain affects the SLE equation. Let $A$ be a simply connected

1019: region connected to the real axis which is at some finite distance

1020: from the origin (see Fig.~\ref{fig:bump}).

1021: %

1022: \begin{figure}

1023: \centering

1024: \includegraphics[width=10cm]{bump.eps}

1025: \caption{\label{fig:bump}\small An SLE hull in ${\bf H}\setminus

1026: A$ and two different ways of removing it: either by first removing

1027: $A$ through $h_0$ and then using a Loewner map $\tilde g_t$ in the

1028: image of ${\bf H}\setminus A$; or by removing $K_t$ first with

1029: $g_t$ and then removing the image of $A$ with $h_t$. Since all

1030: maps are normalised, this diagram commutes.}

1031: \end{figure}

1032: %

1033: Consider a trace $\gamma_t$, with hull $K_t$, which grows from the

1034: origin according to SLE in the domain ${\bf H}\setminus A$.

1035: According to Property 2, we can do this by first making a

1036: conformal mapping $h_0$ which removes $A$, and then a map $\tilde

1037: g_t$ which removes the image $\tilde K_t=h_0(K_t)$. This would be

1038: described by SLE in $h_0({\bf H}\setminus A)$, except that the

1039: Loewner `time' would not in general be the same as $t$. However,

1040: another way to think about this is to first use a SLE map $g_t$ in

1041: $\bf H$ to remove $K_t$, then another map, call it $h_t$, to

1042: remove $g_t(A)$. Since both these procedures end up removing

1043: $K_t\cup A$, and all the maps are assumed to be normalised at

1044: infinity in the standard way (\ref{eqn:asymp}), they must be

1045: identical, that is $h_t\circ g_t=\tilde g_t\circ h_0$ (see

1046: Fig.~\ref{fig:bump}). If $g_t$ maps the growing tip $\tau_t$ to

1047: $a_t$, then after both mappings it goes to $\tilde a_t=h_t(a_t)$.

1048: We would like to understand the law of $\tilde a_t$.

1049: 

1050: Rather than working this out in full generality (see for example

1051: \cite{Wreview}), let us suppose that $A$ is a short vertical

1052: segment $(x,x+i\epsilon)$ with $\epsilon\ll x$, and that $t=dt$ is

1053: infinitesimal. Then, under $g_{dt}$, $x\to x+2dt/x$ and

1054: $\epsilon\to\epsilon(1-2dt/x^2)$. The map that removes this is

1055: (see (\ref{eqn:example}))

1056: \begin{equation}\label{eqn:remove}

1057: h_{dt}(z)=\left((z-x-2dt/x)^2+\epsilon^2(1-2dt/x^2)^2\right)^{1/2}

1058: +x+2dt/x\,.

1059: \end{equation}

1060: In order to find $\tilde a_{dt}$, we need to set

1061: $z=a_{dt}=\sqrt\kappa dB_t$ in this expression. Carefully

1062: expanding this to first order in $dt$, remembering that

1063: $(dB_t)^2=dt$, and also taking the first non-zero contribution in

1064: $\epsilon/x$, gives after a few lines of algebra

1065: \begin{equation}\label{eqn:kappa6}

1066: \tilde a_{dt}=(1-\epsilon^2/x^2)\sqrt\kappa

1067: dB_t+\ffrac12(\kappa-6)(\epsilon^2/x^3)dt\,.

1068: \end{equation}

1069: The factor in front of the stochastic term may be removed by

1070: rescaling $dt$: this restores the correct Loewner time. But there

1071: is also a drift term, corresponding to the effect of $A$. For

1072: $\kappa<6$ we see that the SLE is initially repelled from $A$.

1073: From the point of view of the exploration process for the Ising

1074: model discussed in Sec.~\ref{sec:exploration}, this makes sense:

1075: if the spins along the positive real axis and on $A$ are fixed to

1076: be up, then the spin just above the origin is more likely to be up

1077: than down, and so $\gamma$ is more likely to turn to the left.

1078: 

1079: For $\kappa=6$, however, this is no longer the case: the presence

1080: of $A$ does not affect the initial behaviour of the curve. This is

1081: a particular case of the property of \em locality \em when

1082: $\kappa=6$, which states that, for any $A$ as defined above, the

1083: law of $K_t$ in ${\bf H}\setminus A$ is, up to a time

1084: reparametrisation, the same as the law of $K_t$ in $\bf H$, as

1085: long as $K_t\cap A=\emptyset$. That is, up to the time that the

1086: curve hits $A$, it doesn't know it's there. Such a property would

1087: be expected for the cluster boundaries of uncorrelated Ising spins

1088: on the lattice, i.e.~percolation. This is then consistent with the

1089: identification of percolation cluster boundaries with SLE$_6$.

1090: %

1091: \subsubsection{Restriction}\label{sec:restriction}

1092: %

1093: It is also interesting to work out how the local scale transforms

1094: in going from $a_t$ to $\tilde a_t$. A measure of this is

1095: $h'_t(a_t)$. A similar calculation starting from

1096: (\ref{eqn:remove}) gives, in the same limit as above,

1097: \begin{equation}

1098: d\big(h'(a_t)\big)=

1099: h'_{dt}(a_{dt})-h_0'(0)=(\epsilon^2/x^3)\sqrt\kappa

1100: dB_t+\ffrac12\left(

1101: (\epsilon^4/x^6)+(\kappa-\ffrac83)(3\epsilon^2/x^4)\right)\,.

1102: \end{equation}

1103: Now something special happens when $\kappa=\frac83$. The drift

1104: term in $d\big(h'(a_t)\big)$ does not then vanish, but if we take

1105: the appropriate power $d\big(h'_t(a_t)^{5/8}\big)$ it does. This

1106: implies that the \em mean \em of $h'_t(a_t)^{5/8}$ is conserved.

1107: Now at $t=0$ it takes the value $\Phi_A'(0)^{5/8}$, where

1108: $\Phi_A=h_0$ is the map that removes $A$. If $K_t$ hits $A$ at

1109: time $T$ it can be seen from (\ref{eqn:remove}) that $\lim_{t\to

1110: T}h'_t(a_t)^{5/8}=0$. On the other hand, if it never hits $A$ then

1111: $\lim_{t\to\infty}h'_t(a_t)^{5/8}=1$. Therefore $\Phi_A'(0)^{5/8}$

1112: gives the \em probability that the curve $\gamma$ does not

1113: intersect $A$\em.

1114: 

1115: This is a remarkable result in that it depends only on the value

1116: of $\Phi'_A$ at the starting point of the SLE (assuming of course

1117: that $\Phi_A$ is correctly normalised at infinity.) However it has

1118: the following even more interesting consequence. Let

1119: $\hat\Phi_A(z)=\Phi_A(z)-\Phi(0)$. Consider the ensemble of all

1120: SLE$_{8/3}$ in $\bf H$, and the sub-ensemble consisting of all

1121: those curves $\gamma$ which do not hit $A$. Then the measure on

1122: the image $\hat\Phi_A(\gamma)$ in $\bf H$ is again given by

1123: SLE$_{8/3}$. The way to show this is to realise that the measure

1124: on $\gamma$ is characterised by the probability $P(\gamma\cap

1125: A'=\emptyset)$ that $\gamma$ does not hit $A'$ for all possible

1126: $A'$. The probability that $\hat\Phi_A(\gamma)$ does not hit $A'$,

1127: given that $\gamma$ does not hit $A$, is the ratio of the

1128: probabilities $P(\gamma\cap{\hat\Phi_A}^{-1}(A')=\emptyset)$ and

1129: $P(\gamma\cap A=\emptyset)$. By the above result, the first factor

1130: is the derivative at the origin of the map

1131: $\hat\Phi_{A'}\circ\hat\Phi_A$ which removes $A$ then $A'$, while

1132: the second is the derivative of the map which removes $A$. Thus

1133: \begin{equation}

1134: P(\hat\Phi_A(\gamma)\cap A'=\emptyset|\gamma\cap A=\emptyset)=

1135: \left({(\hat\Phi_{A'}\circ\hat\Phi_A)'(0)\over

1136: {\hat\Phi_A}'(0)}\right)^{5/8}={\hat\Phi_{A'}}'(0)^{5/8}

1137: =P(\gamma\cap A'=\emptyset)\,.

1138: \end{equation}

1139: Since this is true for all $A'$, it follows that the law of

1140: $\hat\Phi_A(\gamma)$ given that $\gamma$ does not intersect $A$ is

1141: the same as that of $\gamma$. This is called the \em restriction

1142: property\em. Note that while, according to Property 2, the law of

1143: an SLE in any simply connected subset of $\bf H$ is determined by

1144: the conformal mapping of this subset to $\bf H$, the restriction

1145: property is stronger than this, and it holds only when

1146: $\kappa=\frac83$.

1147: 

1148: We expect such a property to hold for the continuum limit of

1149: self-avoiding walks, assuming it exists. On the lattice, every

1150: walk of the same length is counted with the same weight. That is,

1151: the measure is uniform. If we consider the sub-ensemble of such

1152: walks which avoid a region $A$, the measure on the remainder

1153: should still be uniform. This will be true if the restriction

1154: property holds. This supports the identification of self-avoiding

1155: walks with SLE$_{8/3}$.

1156: 

1157: 

1158: \subsection{Radial SLE and the winding angle}\label{sec:radialsle}

1159: So far we have discussed a version of SLE that gives a conformally

1160: invariant measure on curves which connect two distinct boundary

1161: points of a simply connected domain $\cal D$. For this reason it

1162: is called \em chordal \em SLE. There are variants which describe

1163: other situations. For example, one could consider curves $\gamma$

1164: which connect a given point $r_1$ on the boundary to an interior

1165: point $r_2$. The Riemann mapping theorem allows us to map

1166: conformally onto another simple connected domain, with $r_2$ being

1167: mapped to any preassigned interior point. It is simplest to choose

1168: for the standard domain the unit disc $\bf U$, with $r_2$ being

1169: mapped to the origin. So we are considering curves $\gamma$ which

1170: connect a given point $e^{i\theta_0}$ on the boundary with the

1171: origin. As before, we may consider growing the curve dynamically.

1172: Let $K_t$ be the hull of that portion which exists up to time $t$.

1173: Then there exists a conformal mapping $g_t$ which takes ${\bf

1174: U}\setminus K_t$ to $\bf U$, such that $g_t(0)=0$. There is one

1175: more free parameter, which corresponds to a global rotation: we

1176: use this to impose the condition that $g_t'(0)$ is real and

1177: positive. One can then show that, as the curve grows, this

1178: quantity is monotonically increasing, and we can use this to

1179: reparametrise time so that $g_t'(0)=e^t$. This normalised mapping

1180: then takes the growing tip $\tau_t$ to a point $e^{i\theta_t}$ on

1181: the boundary.

1182: 

1183: Loewner's theorem tells us that $\dot g_t(z)/g_t(z)$, when

1184: expressed as a function of $g_t(z)$, should be holomorphic in

1185: $\overline{{\bf U}}$ apart from a simple pole at $e^{i\theta_t}$.

1186: Since $g_t$ preserves the unit circle outside $K_t$, $\dot

1187: g_t(z)/g_t(z)$ should be pure imaginary when $|g_t(z)|=1$, and in

1188: order that $g_t'(0)=e^t$, it should approach 1 as $g_t(z)\to0$.

1189: The only possibility is

1190: \begin{equation}\label{eqn:radialsle}

1191: {dg_t(z)\over dt}=-g_t(z)\,{g_t(z)+e^{i\theta_t} \over

1192: g_t(z)-e^{i\theta_t}}\,.

1193: \end{equation}

1194: This is the radial Loewner equation. In fact this is the version

1195: considered by Loewner\cite{Loewner}.

1196: 

1197: It can now be argued, as before, that given Properties 1 \& 2

1198: (suitably reworded to cover the case when $r_2$ is an interior

1199: point) together with reflection, $\theta_t$ must be proportional

1200: to a standard Brownian motion. This defines radial SLE. It is not

1201: immediately obvious how the radial and chordal versions are

1202: related. However, it can be shown that, if the trace of radial SLE

1203: hits the boundary of the unit disc at $e^{i\theta_{t_1}}$ at time

1204: $t_1$, then the law of $K_t$ in radial SLE, for $t<t_1$, is the

1205: same chordal SLE conditioned to begin at $e^{i\theta(0)}$ and end

1206: at $e^{i\theta_{t_1}}$, up to a reparametrisation of time. This

1207: assures us that, in using the chordal and radial versions with the

1208: same $\kappa$, we are describing the same physical problem.

1209: 

1210: However, one feature that the trace of radial SLE possesses which

1211: chordal SLE does not is the property that it can wind around the

1212: origin. The winding angle at time $t$ is simply

1213: $\theta_t-\theta_0$. Therefore it is normally distributed with

1214: variance $\kappa t$. At this point we can make a connection to the

1215: Coulomb gas analysis of the O$(n)$ model in

1216: Sec.~\ref{sec:winding1}, where it was shown that the variance in

1217: the winding angle on a cylinder of length $L$ is asymptotically

1218: $(4/g)L$. A semi-infinite cylinder, parametrised by $w$, is

1219: conformally equivalent to the unit disc by the mapping $z=e^{-w}$.

1220: Asymptotically, ${\rm Re}\,w\to{\rm Re}\,w-t$ under Loewner

1221: evolution. Thus we can identify $L\sim t$ and hence

1222: \begin{equation}\label{eqn:kappag}

1223: \kappa=4/g\,.

1224: \end{equation}

1225: \subsubsection{Identification with lattice

1226: models}\label{sec:identification} This result allows use to make a

1227: tentative identification with the various realisations of the

1228: O$(n)$ model described in Sec.~\ref{sec:Onmodel}. We have, using

1229: (\ref{eqn:kappag}), $n=-2\cos(4\pi/\kappa)$ with

1230: $2\leq\kappa\leq4$ describing the critical point at $x_c$, and

1231: $4<\kappa\leq8$ corresponding to the dense phase. Some important

1232: special cases are therefore:

1233: \begin{itemize}

1234: \item $\kappa=-2$: loop-erased random walks (proven in

1235: \cite{LSWUST}); \item $\kappa=\frac83$: self-avoiding walks, as

1236: already suggested by the restriction property,

1237: Sec.~\ref{sec:restriction}; unproven, but see \cite{LSWSAW} for

1238: many consequences; \item $\kappa=3$: cluster boundaries in the

1239: Ising model, however as yet unproven; \item $\kappa=4$: BCSOS

1240: model of roughening transition (equivalent to the 4-state Potts

1241: model and the double dimer model), as yet unproven; also certain

1242: level lines of a gaussian random field and the `harmonic explorer'

1243: (proven in \cite{SSHE}); also believed to be dual to the

1244: Kosterlitz-Thouless transition in the XY model; \item $\kappa=6$:

1245: cluster boundaries in percolation (proven in \cite{Smirnov});

1246: \item $\kappa=8$: dense phase of self-avoiding walks; boundaries

1247: of uniform spanning trees (proven in \cite{LSWUST}).

1248: \end{itemize}

1249: 

1250: It should be noted that no lattice candidates for $\kappa>8$, or

1251: for the dual values $\kappa<2$, have been proposed. This possibly

1252: has to do with the fact that, for $\kappa>8$, the SLE trace is not

1253: reversible: the law on curves from $r_1$ to $r_2$ is not the same

1254: as the law obtained by interchanging the points. Evidently, curves

1255: in equilibrium lattice models should satisfy reversibility.

1256: %

1257: \newpage

1258: %

1259: \section{Calculating with SLE}\label{sec:calc}

1260: SLE shows that the measure on the continuum limit of single curves

1261: in various lattice models is given in terms of 1d Brownian motion.

1262: However, it is not at all clear how thereby to deduce interesting

1263: physical consequences.  We first describe two relatively simple

1264: computations in two-dimensional percolation which can be done

1265: using SLE.

1266: \subsection{Schramm's formula}\label{sec:schrammformula}

1267: Given a curve $\gamma$ connecting two points $r_1$ and $r_2$ on

1268: the boundary of a domain $\cal D$, what is the probability that it

1269: passes to the left of a given interior point? This is not a

1270: question which is natural in conventional approaches to critical

1271: behaviour, but which is very simply answered within

1272: SLE\cite{Schrammformula}.

1273: 

1274: As usual, we can consider $\cal D$ to be the upper half plane, and

1275: take $r_1=a_0$ and $r_2$ to be at infinity. The curve is then

1276: described by chordal SLE starting at $a_0$. Label the interior

1277: point by the complex number $\zeta$.

1278: 

1279: Denote the probability that $\gamma$ passes to the left of $\zeta$

1280: by $P(\zeta,\bar\zeta;a_0)$ (we include the dependence on

1281: $\bar\zeta$ to emphasise the fact that this is a not a holomorphic

1282: function.) Consider evolving the SLE for an infinitesimal time

1283: $dt$. The function $g_{dt}$ will map the remainder of $\gamma$

1284: into its image $\gamma'$, which, however, by Properties 1 \& 2,

1285: will have the same measure as SLE started from

1286: $a_{dt}=a_0+\sqrt\kappa dB_t$. At the same time, $\zeta\to

1287: g_{dt}(\zeta)=\zeta+2dt/(\zeta-a_0)$. Moreover, $\gamma'$ lies to

1288: the left of $\zeta'$ iff $\gamma$ lies to the left of $\zeta$.

1289: Therefore

1290: \begin{equation}

1291: P\big(\zeta,\bar\zeta;a_0\big)= \langle

1292: P\big(\zeta+2dt/(\zeta-a_0),\bar\zeta+2dt/(\bar\zeta-a_0),a_0

1293: +\sqrt\kappa dB_t\big)\rangle\,,

1294: \end{equation}

1295: where the average $\langle\ldots\rangle$\ is over all realisations

1296: of Brownian motion $dB_t$ up to time $dt$. Taylor expanding, using

1297: $\langle dB_t\rangle=0$ and $\langle(dB_t)^2\rangle=dt$, and

1298: equating the coefficient of $dt$ to zero gives

1299: \begin{equation}\label{eqn:pde1}

1300: \left({2\over\zeta-a_0}{\partial\over\partial\zeta}

1301: +{2\over\bar\zeta-a_0}{\partial\over\partial\bar\zeta}

1302: +\frac\kappa 2{\partial^2\over\partial a_0^2}\right)

1303: P\big(\zeta,\bar\zeta;a_0\big)=0\,.

1304: \end{equation}

1305: Thus $P$ satisfies a linear second-order partial differential equation,

1306: typical of conditional probabilities in stochastic differential equations.

1307: 

1308: By scale invariance $P$ in fact depends only on the angle $\theta$

1309: subtended between $\zeta-a_0$ and the real axis. Thus

1310: (\ref{eqn:pde1}) reduces to an ordinary second-order linear

1311: differential equation, which is in fact hypergeometric. The

1312: boundary conditions are that $P=0$ and $1$ when $\theta=\pi$ and

1313: $0$ respectively, which gives (specialising to $\kappa=6$)

1314: \begin{equation}\label{eqn:schrammformula}

1315: P=\frac12+{\Gamma(\ffrac23)\over\sqrt\pi\Gamma(\ffrac16)}

1316: (\cot\theta) {}_2F_1(\ffrac12,\ffrac23,\ffrac32;-\cot^2\theta)\,.

1317: \end{equation}

1318: Note that this may also be written in terms of a single quadrature

1319: since one solution of (\ref{eqn:pde1}) is $P=$ const.

1320: %

1321: \subsection{Crossing probability}

1322: Given a critical percolation problem inside a simply connected

1323: domain $\cal D$, what is the probability that a cluster connects

1324: two disjoint segments $AB$ and $CD$ of the boundary? This problem

1325: was conjectured to be conformally invariant and (probably) first

1326: studied numerically in \cite{Langlandsetal}. A formula based on

1327: CFT as well as a certain amount of guesswork was conjectured in

1328: \cite{JCcrossing}. It was proved, for the continuum limit of site

1329: percolation on the triangular lattice, by Smirnov\cite{Smirnov}.

1330: 

1331: Within SLE, it takes a certain amount of ingenuity\cite{Schramm}

1332: to relate this problem to a question about a single curve. As

1333: usual, let $\cal D$ be the upper half plane. It is always possible

1334: to make a fractional linear conformal mapping which takes $AB$

1335: into $(-\infty,x_1)$ and $CD$ into $(0,x_2)$, where $x_1<0$ and

1336: $x_2>0$. Now go back to the lattice picture and consider critical

1337: site percolation on the triangular lattice in the upper half

1338: plane, so that each site is independently coloured black or white

1339: with equal probabilities $\frac12$. Choose all the boundary sites

1340: on the positive real axis to be white, all those on the negative

1341: real axis to be black (see Fig.~\ref{fig:crossing}). There is a

1342: cluster boundary starting at the origin, which, in the continuum

1343: limit, will be described by SLE$_6$. Since $\kappa>4$, it

1344: repeatedly hits the real axis, both to the L and R of the origin.

1345: Eventually every point on the real axis is swallowed. Either $x_1$

1346: is swallowed before $x_2$, or vice versa.

1347: %

1348: \begin{figure}

1349: \centering

1350: \includegraphics[width=8cm]{crossing.eps}

1351: \caption{\label{fig:crossing}\small Is there a crossing on the

1352: white discs from $(0,x_2)$ to $(-\infty,x_1)$? This happens if and

1353: only if $x_1$ gets swallowed by the SLE before $x_2$.}

1354: \end{figure}

1355: %

1356: 

1357: Note that every site on the L of the curve is black, and every

1358: site on its R is white. Suppose that $x_1$ is swallowed before

1359: $x_2$. Then, at the moment it is swallowed, there exists a

1360: continuous path on the white sites, just to the R of the curve,

1361: which connects $(0,x_2)$ to the row just above $(-\infty,x_1)$. On

1362: the other hand, if $x_2$ is swallowed before $x_1$, there exists a

1363: continuous path on the black sites, just to the L of the curve,

1364: connecting $0-$ to a point on the real axis to the R of $x_2$.

1365: This path forms a barrier (as in the game of Hex) to the

1366: possibility of a white crossing from $(0,x_2)$ to $(-\infty,x_1)$.

1367: Hence there is such a crossing if and only if $x_1$ is swallowed

1368: before $x_2$ by the SLE curve.

1369: 

1370: Recall that in Sec.~\ref{sec:phases} we related the swallowing of

1371: a point $x_0$ on the real axis to the vanishing of

1372: $x_t=g_t(x_t)-a_t$, which undergoes a Bessel process on the real

1373: line. Therefore

1374: \begin{equation}

1375: {\rm Pr}\big({\rm crossing\ from\ }(0,x_2)\ {\rm to\

1376: }(-\infty,x_1)\big) ={\rm Pr}\big(x_{1t}\ {\rm  vanishes\ before\

1377: } x_{2t}\big)\,.

1378: \end{equation}

1379: 

1380: Denote this by $P\big(x_1,x_2\big)$. By generalising the SLE to

1381: start at $a_0$ rather than $0$, we can write a differential

1382: equation for this in similar manner to (\ref{eqn:pde1}):

1383: \begin{equation}\label{eqn:pde2}

1384: \left({2\over x_1-a_0}{\partial\over\partial x_1} +{2\over

1385: x_2-a_0}{\partial\over\partial x_2} +\frac\kappa

1386: 2{\partial^2\over\partial a_0^2}\right)\

1387: P\big(x_1,x_2;a_0\big)=0\,.

1388: \end{equation}

1389: Translational invariance implies that we can replace

1390: $\partial_{a_0}$ by $-(\partial_{x_1}+\partial_{x_2})$. Finally,

1391: $P$ can in fact depend only on the ratio

1392: $\eta=(x_2-a_0)/(a_0-x_1)$, which again reduces the equation to

1393: hypergeometric form. The solution is (specialising to $\kappa=6$

1394: for percolation)

1395: \begin{equation}\label{eqn:crossing}

1396: P={\Gamma(\ffrac23)\over\Gamma(\ffrac43)\Gamma(\ffrac13)}\eta^{1/3}

1397: {}_2F_1(\ffrac13,\ffrac23,\ffrac43;\eta)\,.

1398: \end{equation}

1399: It should be mentioned that this is but one of a number of

1400: percolation crossing formulae. Another, conjectured by

1401: Watts\cite{Watts}, for the probability that there is cluster which

1402: simultaneously connects $AB$ to $CD$ and $BC$ to $DA$, has since

1403: been proved by Dub\'edat\cite{Dubedat}. However, other formulae,

1404: for example for the mean number of distinct clusters connecting

1405: $AB$ and $CD$\cite{JCmean}, and for the probability that exactly

1406: $N$ distinct clusters cross an annulus\cite{JCannulus}, are as yet

1407: unproven using SLE methods.

1408: %

1409: \subsection{Critical exponents from SLE}\label{sec:sleexponents}

1410: Many of the critical exponents which have previously been

1411: conjectured by Coulomb gas or CFT methods may be derived

1412: rigorously using SLE, once the underlying postulates are assumed

1413: or proved. However SLE describes the measure on just a single

1414: curve, and in the papers of LSW a great deal of ingenuity has gone

1415: into showing how to relate this to all the other exponents. There

1416: is not space in this article to do these justice. Instead we

1417: describe three examples which give the flavour of the arguments,

1418: which initially may appear quite unconventional compared with the

1419: more traditional approaches.

1420: 

1421: \subsubsection{The fractal dimension of

1422: SLE}\label{sec:fractaldim} The fractal dimension of any

1423: geometrical object embedded in the plane can be defined roughly as

1424: follows: let $N(\epsilon)$ be the minimum number of small discs of

1425: radius $\epsilon$ required to cover the object. Then if

1426: $N(\epsilon)\sim\epsilon^{-d_f}$ as $\epsilon\to0$, $d_f$ is the

1427: fractal dimension.

1428: 

1429: One way of computing $d_f$ for a random curve $\gamma$ in the

1430: plane is to ask for the probability $P(x,y,\epsilon)$ that a given

1431: point $\zeta=x+iy$ lies within a distance $\epsilon$ of $\gamma$.

1432: A simple scaling argument shows that if $P$ behaves like

1433: $\epsilon^\delta f(x,y)$ as $\epsilon\to0$, then $\delta=2-d_f$.

1434: We can derive a differential equation for $P$ along the lines of

1435: the previous calculation. The only difference is that under the

1436: conformal mapping $g_{dt}$, $\epsilon\to

1437: |g_{dt}'(\zeta)|\epsilon\sim\big(1-2dt\,{\rm

1438: Re}(1/\zeta^2)\big)\epsilon$. The differential equation (written

1439: for convenience in cartesian coordinates) is

1440: \begin{equation}\label{eqn:pde3}

1441: \left({2x\over x^2+y^2}{\partial\over\partial x}-{2y\over

1442: x^2+y^2}{\partial\over\partial y} +\frac\kappa

1443: 2{\partial^2\over\partial x^2}-{2(x^2-y^2)\over(x^2+y^2)^2}

1444: \epsilon{\partial\over\partial\epsilon}\right)P=0\,.

1445: \end{equation}

1446: Now $P$ is dimensionless and therefore should have the form

1447: $(\epsilon/r)^{2-d_f}$ times a function of the polar angle

1448: $\theta$. In fact, the simple ansatz

1449: $P=\epsilon^{2-d_f}y^\alpha(x^2+y^2)^\beta$, with $\alpha+2\beta=

1450: d_f-2$ satisfies the equation. [The reason this works is connected

1451: with the simple form for the correlator

1452: $\langle\Phi_2\phi_{2,1}\phi_{2,1}\rangle$ discussed in

1453: Sec.~\ref{sec:CS}.] This gives $\alpha=(\kappa-8)^2/8\kappa$,

1454: $\beta=(\kappa-8)/2\kappa$ and

1455: \begin{equation}

1456: d_f=1+\kappa/8\,.

1457: \end{equation}

1458: This is correct for $\kappa\leq8$: otherwise there is another

1459: solution with $\alpha=\beta=0$ and $d_f=2$. A more careful

1460: statement and proof of this result can be found in \cite{Beffara}.

1461: 

1462: We see that the fractal dimension increases steadily from the

1463: value 1 when $\kappa=0$ (corresponding to a straight line) to a

1464: maximum value of 2 when $\kappa=8$. Beyond this value $\gamma$

1465: becomes space-filling: every point in the upper half plane lies on

1466: the curve.

1467: %

1468: \subsubsection{Crossing exponent}\label{sec:crossingexponent}

1469: Consider a critical percolation problem in the upper half plane.

1470: What is the asymptotic behaviour as $r\to\infty$ of the

1471: probability that the interval $(0,1)$ on the real axis is

1472: connected to the interval $(r,\infty)$? We expect this to decay as

1473: some power of $r$. The value of this exponent may be found by

1474: taking the appropriate limit of the crossing formula

1475: (\ref{eqn:crossing}), but instead we shall compute it directly. In

1476: order for there to be a crossing cluster, there must be two

1477: cluster boundaries which also cross between the two intervals, and

1478: which bound this cluster above and below. Denote the upper

1479: boundary by $\gamma$. Then we need to know the probability $P(r)$

1480: of there being another spanning curve lying between $\gamma$ and

1481: $(1,r)$, averaged over all realisations of $\gamma$. Because of

1482: the locality property, the measure on $\gamma$ is independent of

1483: the existence of the lower boundary, and is given by SLE$_6$

1484: conditioned not to hit the real axis along $(1,r)$. Note that

1485: because $\kappa>4$ it will eventually hit the real axis at some

1486: point to the right of $r$. For this reason we can do the

1487: computation for general $\kappa>4$, although it gives the actual

1488: crossing exponent only if $\kappa=6$.

1489: 

1490: Consider the behaviour of $P(r)$ under the conformal mapping $\hat

1491: g_{dt}(z)\sim z+(2dt/z)-\sqrt\kappa dB_t$ (which maps the growing

1492: tip $\tau_t$ into $0$.) The crossing probability should be

1493: conformally invariant and depend only on the ratio of the lengths

1494: of the two intervals, hence, by an argument which by now should be

1495: familiar,

1496: \begin{equation}

1497: P(r)=\langle P\big(\hat g_{dt}(r)/\hat g_{dt}(1)\big)\rangle\,.

1498: \end{equation}

1499: Expanding this out, remembering as usual that $(dB_t)^2=dt$, and

1500: setting to zero the $O(dt)$ term, we find for $r\gg1$

1501: \begin{equation}

1502: (\kappa-2)rP'(r)+\ffrac12\kappa r^2P''(r)=0\,,

1503: \end{equation}

1504: with the solution $P(r)\propto r^{-(\kappa-4)/\kappa}$ for

1505: $\kappa>4$. Setting $\kappa=6$ then gives the result $\frac13$.

1506: 

1507: \subsubsection{The one-arm exponent}\label{sec:onearm}

1508: Consider critical lattice percolation inside some finite region

1509: (for example a disc of radius $R$). What is the probability that a

1510: given site (e.g.~the origin) is connected to a finite segment $S$

1511: of the boundary? This should decay like $R^{-\lambda}$, where

1512: $\lambda$ is sometimes called the one-arm exponent. If we try to

1513: formulate this in the continuum, we immediately run up against the

1514: problem that all clusters are fractal with dimension $<2$, and so

1515: the probability of any given point being in any given cluster is

1516: zero. Instead, one may ask about the probability $P(r)$ that the

1517: cluster connected to $S$ gets within a distance $r$ of the origin.

1518: This should behave like $(r/R)^{\lambda}$. We can now set $R=1$

1519: and treat the problem using radial SLE$_6$.

1520: 

1521: Consider now a radial SLE$_\kappa$ which starts at

1522: $e^{i\theta_0}$. If $\kappa>4$ it will continually hit the

1523: boundary. Let $P(\theta-\theta_0,t)$ be the probability that the

1524: segment $(\theta_0,\theta)$ of the boundary has not been swallowed

1525: by time $t$. Then, by considering the evolution as usual under

1526: $g_{dt}$,

1527: \begin{equation}

1528: P(\theta,\theta_0,t)=\langle

1529: P(\theta+d\theta,\theta_0+d\theta_0,t-dt)\rangle\,,

1530: \end{equation}

1531: where $d\theta=\cot\big((\theta-\theta_0)/2\big)dt$ and

1532: $d\theta_0=\sqrt\kappa dB_t$. Setting $\theta_0=0$ and equating to

1533: zero the $O(dt)$ term, we find the time-dependent differential

1534: equation

1535: \begin{equation}\label{eqn:pde4}

1536: \partial_tP=\cot(\theta/2)\partial_\theta P+

1537: \ffrac12\kappa\partial^2_\theta P\,.

1538: \end{equation}

1539: This has the form of a backwards Fokker-Plank equation.

1540: 

1541: Now, since $g_t'(0)=e^t$, it is reasonable that, after time $t$,

1542: the SLE gets within a distance $O(e^{-t})$ of the origin.

1543: Therefore we can interpret $P$ as roughly the probability that the

1544: cluster connected to $(0,\theta)$ gets within a distance $r\sim

1545: e^{-t}$ of the origin. A more careful argument \cite{LSW1arm}

1546: confirms this. The boundary conditions are $P(0,t)=0$ as

1547: $\theta\to0$, and (with more difficulty) $\partial_\theta

1548: P(\theta,t)=0$ at $\theta=2\pi$. The solution may then be found by

1549: inspection to be

1550: \begin{equation}

1551: P\propto e^{-\lambda t}\big(\sin(\theta/4)\big)^{1-4/\kappa}\,,

1552: \end{equation}

1553: where $\lambda=(\kappa^2-16)/32\kappa$. For percolation this gives

1554: $\frac5{48}$, in agreement with Coulomb gas arguments

1555: \cite{Nienhuis}.

1556: 

1557: The appearance of differential operators such as that in

1558: (\ref{eqn:pde4}) will become clear from the CFT perspective in

1559: Sec.~\ref{sec:CS}. If instead of choosing Neuman boundary

1560: conditions at $\theta=2\pi$ we impose $P=0$, the same equation

1561: gives the bulk 2-leg exponent $x_2$, which is also related to the

1562: fractal dimension by $d_f=2-x_2$.

1563: %

1564: \newpage

1565: %

1566: \section{Relation to conformal field theory}\label{sec:cft}

1567: \subsection{Basics of CFT.}\label{sec:basics}

1568: The reader who already knows a little about CFT will have

1569: recognised the differential equations in Sec.~\ref{sec:calc} as

1570: being very similar in form to the BPZ equations\cite{BPZ}

1571: satisfied by the correlation functions of a $\phi_{2,1}$ operator,

1572: corresponding to a highest weight representation of the Virasoro

1573: algebra with a level 2 null state.

1574: 

1575: For those readers for whom the above paragraph makes no sense, and

1576: in any case to make the argument self-contained, we first

1577: introduce the concepts of (boundary) conformal field theory

1578: (BCFT.) We stress that these are heuristic in nature -- they serve

1579: only as a guide to formulating the basic principles of CFT which

1580: can then be developed into a mathematically consistent theory. For

1581: a longer introduction to BCFT see \cite{JCBCFT} and, for a

1582: complete account of CFT, \cite{YellowBook}.

1583: 

1584: We have at the back of our minds a euclidean field theory defined

1585: as a path integral over some set of fundamental fields

1586: $\{\psi(r)\}$. The partition function is $Z=\int

1587: e^{-S(\{\psi\})}[d\psi]$ where the action $S(\{\psi\})=\int_{\cal

1588: D}{\cal L}(\{\psi\})d^2\!r$ is an integral over a local lagrangian

1589: density. These fields may be thought of as smeared-out continuum

1590: versions of the lattice degrees of freedom. As in any field

1591: theory, this continuum limit involves renormalisation. There are

1592: so-called local scaling operators $\phi_j^{(0)}(r)$ which are

1593: particular functionals of the fundamental degrees of freedom,

1594: which have the property that we can define renormalised scaling

1595: operators $\phi_j(r)=a^{-x_j}\phi_j^{(0)}(r)$ whose correlators

1596: are finite in the continuum limit $a\to0$, that is

1597: \begin{equation}

1598: \lim_{a\to0}a^{-\sum_jx_j}\langle\phi_1^{(0)}(r_1)\ldots

1599: \phi_N^{(0)}(r_N)\rangle=\langle\phi_1(r_1)\ldots

1600: \phi_N(r_N)\rangle

1601: \end{equation}

1602: exists. The numbers $x_j$ are called the scaling dimensions, and

1603: are related to the various critical exponents. They are related to

1604: the conformal weights $(h_j,\bar h_j)$ by $x_j=h_j+\bar h_j$; the

1605: difference $h_j-\bar h_j=s_j$ is called the spin of $\phi_j$, and

1606: describes its behaviour under rotations. There are also \em

1607: boundary operators\em, localised on the boundary, which have only

1608: a single conformal weight equal to their scaling dimension.

1609: 

1610: The theory is developed independently of any particular set of

1611: fundamental fields or lagrangian. An important role in this is

1612: played by the stress tensor $T^{\mu\nu}(r)$, defined as the local

1613: response of the action to a change in the metric:

1614: \begin{equation}

1615: \delta S=(1/4\pi)\int_{\cal D}T^{\mu\nu}\delta g_{\mu\nu}d^2r\,.

1616: \end{equation}

1617: Invariance under local rotations and scale transformations usually

1618: implies that $T^{\mu\nu}$ is symmetric and traceless:

1619: $T^\mu_\mu=0$. This also implies invariance under conformal

1620: transformations, corresponding to $\delta g_{\mu\nu}\propto

1621: f(r)g_{\mu\nu}$.

1622: 

1623: In two dimensional flat space, infinitesimal coordinate

1624: transformations $r^\mu\to {r'}^\mu=r^\mu+\alpha^\mu(r)$ correspond

1625: to infinitesimal transformations of the metric with $\delta

1626: g^{\mu\nu}=-(\partial^\mu\alpha^\mu+

1627: \partial^\nu\alpha^\nu)$. It is important to

1628: realise that under these transformations the underlying lattice,

1629: or UV cut-off, is not transformed. Otherwise they would amount to

1630: a trivial reparametrisation. For a conformal transformation,

1631: $\alpha^\mu(r)$ is given by an analytic function: in complex

1632: coordinates $(z,\bar z)$, $\partial_{\bar z}\alpha^z =0$, so

1633: $\alpha^z\equiv\alpha(z)$ is holomorphic. However, such a function

1634: cannot be small everywhere (unless it is constant), so it is

1635: necessary to consider coordinate transformations which are not

1636: everywhere conformal.

1637: 

1638: Consider therefore two concentric semicircles $\Gamma_1$ and $\Gamma_2$

1639: in the upper half plane, centred on the origin, and of radii $R_1<R_2$.

1640: For $|r|<R_1$ let $\alpha^\mu$ be conformal, with $\alpha^z=\alpha(z)$,

1641: while for $|r|>R_2$ take $\alpha^\mu=0$. In between,

1642: $\alpha^\mu$ is not conformal, but is differentiable,

1643: so that

1644: $\delta S=(-1/2\pi)\int_{R_1<|r|<R_2}T^{\mu\nu}\partial_\mu\alpha_\nu d^2\!r$.

1645: This can be integrated by parts to give a term

1646: $(1/2\pi)\int_{R_1<|r|<R_2}\partial_\mu T^{\mu\nu}\alpha_\nu d^2\!r$ (which

1647: must vanish because $\alpha_\nu$ is arbitrary in this region, implying

1648: that $\partial_\mu T^{\mu\nu}=0$) and two surface terms. That on $\Gamma_2$

1649: vanishes because $\alpha^\mu=0$ there. We are left with

1650: \begin{equation}\label{eqn:deltaS}

1651: \delta S=(1/2\pi)\int_{\Gamma_1}T^{\mu\nu}\alpha_\mu

1652: \epsilon^{\nu\lambda}d\ell_\lambda\,,

1653: \end{equation}

1654: where $d\ell^\lambda$ is the line element along $\Gamma_1$.

1655: 

1656: The fact that $T^{\mu\nu}$ is conserved means, in complex

1657: coordinates, that $\partial_{\bar z}T_{zz}=\partial_zT_{\bar z\bar

1658: z}=0$, so that the correlations functions of $T(z)\equiv T_{zz}$

1659: are holomorphic functions of $z$, while those of $\overline

1660: T\equiv T_{\bar z\bar z}$ are antiholomorphic. (\ref{eqn:deltaS})

1661: may then be written

1662: \begin{equation}\label{eqn:deltaS2}

1663: \delta S=(1/2\pi i)\int_{\Gamma_1}T(z)\alpha(z)dz+{\rm c.c.}\,.

1664: \end{equation}

1665: 

1666: In any field theory with a boundary, it is necessary to impose

1667: some boundary condition. It can be argued that any translationally

1668: invariant boundary condition flows under the RG to conditions

1669: satisfying $T_{xy}=0$, which in complex coordinates means that

1670: $T=\overline T$ on the real axis. This means that the correlators

1671: of $\overline T$ are those of $T$ analytically continued into the

1672: lower half plane. The second term in (\ref{eqn:deltaS2}) may then

1673: be dropped if the contour in the first term is around a complete

1674: circle.

1675: 

1676: The conclusion of all this is that the effect of an infinitesimal

1677: conformal transformation on any correlator of observables inside

1678: $\Gamma_1$ is the same as inserting a contour integral $\int

1679: T(z)\alpha(z)dz/2\pi i$ into the correlator.

1680: 

1681: Another important element of CFT is the operator product expansion

1682: (OPE) of the stress tensor with other local operators. Since $T$

1683: is holomorphic, this has the form

1684: \begin{equation}\label{eqn:Tphi}

1685: T(z)\cdot\phi(0)=\sum_n z^{-n-2}\phi^{(n)}(0)\,,

1686: \end{equation}

1687: where the $\phi^{(n)}$ are (possibly new) local operators. By

1688: taking $\alpha(z)\propto z$ (corresponding to a scale

1689: transformation) we see that $\phi^{(0)}=h\phi$, where $h$ is its

1690: scaling dimension. Similarly, by taking $\alpha=$ const.,

1691: $\phi^{(-1)}=\partial_x\phi$. Local operators for which

1692: $\phi^{(n)}$ vanishes for $n\geq1$ are called primary. $T$ itself

1693: is not primary: its OPE with itself takes the form

1694: \begin{equation}\label{eqn:TT}

1695: T(z)\cdot T(0)=c/2z^4+(2/z^2)T(0)+(1/z)\partial_zT(0)+\cdots\,,

1696: \end{equation}

1697: where $c$ is the conformal anomaly number, ubiquitous in CFT. For

1698: example, the partition function on a long cylinder of length $L$

1699: and circumference $\ell$ behaves as $\exp(\pi cL/\ell)$,

1700: cf.~(\ref{eqn:c6}).

1701: %

1702: \subsection{Radial quantisation}

1703: This is the most important concept in understanding the link

1704: between SLE and CFT. We introduce it in the context of boundary

1705: CFT. As before, suppose there is some set of fundamental fields

1706: $\{\psi(r)\}$, with a Gibbs measure $e^{-S[ \psi]}[d\psi]$. Let

1707: $\Gamma$ be a semicircle in the upper half plane, centered on the

1708: origin. The Hilbert space of the BCFT is the function space (with

1709: a suitable norm) of field configurations $\{\psi_\Gamma\}$ on

1710: $\Gamma$.

1711: 

1712: The vacuum state is given by weighting each state

1713: $|\psi'_\Gamma\rangle$ by the (normalised) path integral

1714: restricted to the interior of $\Gamma$ and conditioned to take the

1715: specified values $\psi'_\Gamma$ on the boundary:

1716: \begin{equation}

1717: |0\rangle=\int[d\psi'_\Gamma]\int_{\psi_\Gamma=\psi'_\Gamma}

1718: [d\psi]\,e^{-S[\psi]}\,|\psi'_\Gamma\rangle\,.

1719: \end{equation}

1720: Note that because of scale invariance different choices of the

1721: radius of $\Gamma$ are equivalent, up to a normalisation factor.

1722: 

1723: Similarly, inserting a local operator $\phi(0)$ at the origin into

1724: the path integral defines a state $|\phi\rangle$. This is called

1725: the operator-state correspondence of CFT. If we also insert

1726: $(1/2\pi i)\int_Cz^{n+1}T(z)dz$, where $C$ lies inside $\Gamma$,

1727: we get a state $L_n|\phi\rangle$. The $L_n$ act linearly on the

1728: Hilbert space. From the OPE (\ref{eqn:Tphi}) we see that

1729: $L_n|\phi\rangle\propto|\phi^{(n)}\rangle$, and that, in

1730: particular, $L_0|\phi\rangle=h_\phi|\phi\rangle$. If $\phi$ is

1731: primary, $L_n|\phi\rangle=0$ for $n\geq1$. From the OPE

1732: (\ref{eqn:TT}) of $T$ with itself follow the commutation relations

1733: for the $L_n$

1734: \begin{equation}\label{eqn:vir}

1735: [L_n,L_m]=(n-m)L_{n+m}+\ffrac1{12}cn(n^2-1)\delta_{n,-m}\,,

1736: \end{equation}

1737: which are known as the Virasoro algebra. The state $|\phi\rangle$

1738: together with all its descendants, formed by acting on

1739: $|\phi\rangle$ an arbitrary number of times with the $L_n$ with

1740: $n\leq-1$, give a highest weight representation (where the weight

1741: is defined as the eigenvalue of $-L_0$.)

1742: 

1743: There is another way of generating such a highest weight

1744: representation. Suppose the boundary conditions on the negative

1745: and positive real axes are both conformal, that is they satisfy

1746: $T=\overline T$, but they are different. The vacuum with these

1747: boundary conditions gives a highest weight state which it is

1748: sometimes useful to think of as corresponding to the insertion of

1749: a `boundary condition changing' (bcc) operator at the origin. An

1750: example is the continuum limit of an Ising model in which the

1751: spins on the negative real axis are $-1$, and those on the

1752: positive axis are $+1$.

1753: 

1754: \subsection{Curves and states.}\label{sec:curves} In this section

1755: we describe a way of associating states in the Hilbert space of

1756: the BCFT with the growing curves of the Loewner process. This was

1757: first understood by M.~Bauer and D.~Bernard\cite{BB}, but we shall

1758: present the arguments slightly differently.

1759: 

1760: The boundary conditions associated with a bcc operator guarantee

1761: the existence, on the lattice, of a domain wall connecting the

1762: origin to infinity. Given a particular realisation $\gamma$, we

1763: can condition the Ising spins on its existence. We would like to

1764: be able to assume that this property continues to hold in the

1765: continuum limit: that is, we can condition the fields $\{\psi\}$

1766: on the existence of a such a curve. However, this involves

1767: conditioning on an event with probability zero: it turns out that

1768: in general the probability that, with respect to the measure in

1769: the path integral, the probability that a domain wall hits the

1770: real axis somewhere in an interval of length $\epsilon$ vanishes

1771: like $\epsilon^h$. In what follows we shall regard $\epsilon$ as

1772: small but fixed, and assume that the usual properties of SLE are

1773: applicable to this more general case.

1774: 

1775: Any such curve may be generated by a Loewner process: denote as

1776: before the part of the curve up to time $t$ by $\gamma_t$. The

1777: existence of this curve depends on only the field configurations

1778: $\psi$ in the interior of $\Gamma$, as long as $\gamma_t$ lies

1779: wholly inside this region. Then we can condition the fields

1780: contributing to the path integral on the existence of $\gamma_t$,

1781: thus defining a state

1782: \begin{equation}

1783: |\gamma_t\rangle =

1784: \int[d\psi'_\Gamma]\int_{\psi_\Gamma=\psi'_\Gamma;\gamma_t}

1785: [d\psi]\,e^{-S[\psi]}\,|\psi'_\Gamma\rangle\,.

1786: \end{equation}

1787: The path integral (over the whole of the upper half plane, not

1788: just the interior of $\Gamma$), when conditioned on $\gamma_t$,

1789: gives a measure $d\mu(\gamma_t)$ on these curves. The state

1790: \begin{equation}\label{eqn:ht}

1791: |h\rangle=|h_t\rangle\equiv\int d\mu(\gamma_t)|\gamma_t\rangle

1792: \end{equation}

1793: is in fact independent of $t$, since it is just given by the path

1794: integral conditioned on there being a curve connecting the origin

1795: to infinity, as guaranteed by the boundary conditions. In fact we

1796: see that $|h\rangle$ is just the state corresponding to a boundary

1797: condition changing operator at the origin.

1798: 

1799: However, $d\mu(\gamma_t)$ is also given by the measure on $a_t$ in

1800: Loewner evolution, through the iterated sequence of conformal

1801: mappings satisfying $d\hat g_t=2dt/\hat g_t-da_t$. This

1802: corresponds to an infinitesimal conformal mapping of the upper

1803: half plane minus $K_t$. As explained in the previous section,

1804: $d\hat g_t$ corresponds to inserting $(1/2\pi

1805: i)\int_C(2dt/z-da_t)T(z)dz$. In operator language, this

1806: corresponds to acting on $|\gamma_t\rangle$ with

1807: $2L_{-2}dt-L_{-1}da_t$ where $L_n=(1/2\pi i)\int_Cz^{n+1}T(z)dz$.

1808: Thus, for any $t_1<t$,

1809: \begin{equation}

1810: |g_{t_1}(\gamma_t)\rangle ={\bf

1811: T}\exp\left(\int_0^{t_1}\big(2L_{-2}dt'-L_{-1}da_{t'}\big)\right)

1812: |\gamma_t\rangle\,,

1813: \end{equation}

1814: where $\bf T$ denotes a time-ordered exponential.

1815: 

1816: The measure on $\gamma_t$ is the product of the measure of

1817: $\gamma_t\setminus\gamma_{t_1}$, conditioned on $\gamma_{t_1}$,

1818: with the unconditioned measure on $\gamma_{t_1}$. The first is the

1819: same as the unconditioned measure on $g_{t_1}(\gamma_t)$, and the

1820: second is given by the measure on $a_{t'}$ for $t'\in[0,t_1]$.

1821: Thus we can rewrite both the measure and the state in

1822: (\ref{eqn:ht}) as

1823: \begin{equation}

1824: |h_t\rangle=\int d\mu(g_{t_1}(\gamma_t))\int

1825: d\mu(a_{t';t'\in[0,t_1]}) {\bf

1826: T}e^{\int_{t_1}^0\big(2L_{-2}dt'-L_{-1}da_{t'}\big)}

1827: |g_{t_1}(\gamma_t)\rangle\,.

1828: \end{equation}

1829: For SLE, $a_t$ is proportional to a Brownian process. The

1830: integration over realisations of this for $t'\in[0,t_1]$ may be

1831: performed by breaking up the time interval into small segments of

1832: size $\delta t$, expanding out the exponential to $O(\delta t)$,

1833: using $(B_{\delta t})^2\approx \delta t$, and re-exponentiating.

1834: The result is

1835: \begin{equation}

1836: |h_t\rangle=\exp\left(-\big(2L_{-2}-(\kappa/2)L_{-1}^2\big)t_1\right)

1837: |h_{t-t_1}\rangle\,.

1838: \end{equation}

1839: But, as we argued earlier, $|h_t\rangle$ is independent of $t$,

1840: and therefore

1841: \begin{equation}

1842: \label{l2l1} \big(2L_{-2}-(\kappa/2)L_{-1}^2\big)|h\rangle=0\,.

1843: \end{equation}

1844: 

1845: This means that the descendant states $L_{-2}|h\rangle$ and

1846: $L^2_{-1}|h\rangle$ are linearly dependent. We say that the

1847: Virasoro representation corresponding to $|h\rangle$ has a null

1848: state at level 2. From this follow an number of important

1849: consequences. Acting on (\ref{l2l1}) with $L_1$ and $L_2$, and

1850: using the Virasoro algebra (\ref{eqn:vir}) and the fact that

1851: $L_1|h\rangle=L_2|h\rangle=0$ while $L_0|h\rangle=h|h\rangle$,

1852: leads to

1853: \begin{eqnarray}

1854: h&=&h_{2,1}={6-\kappa\over2\kappa}\,;\\

1855: c&=&{(3\kappa-8)(6-\kappa)\over2\kappa}\,.

1856: \end{eqnarray}

1857: These are the fundamental relations between the parameter $\kappa$

1858: of SLE and the data of CFT. The conventional notation $h_{2,1}$

1859: comes from the Kac formula in CFT which we do not need here. In

1860: fact this is appropriate to the case $\kappa<4$: for $\kappa>4$ it

1861: corresponds to $h_{1,2}$. (To further confuse the matter, some

1862: authors reverse the labels.) Note that the boundary exponent $h$

1863: parametrises the failure of locality in (\ref{eqn:kappa6}). From

1864: CFT we may also deduce that, with respect to the path integral

1865: measure, the probability that a curve connects small intervals of

1866: size $\epsilon$ about points $r_1$, $r_2$ on the real axis behaves

1867: like

1868: \begin{equation}

1869: \epsilon^{2h_{2,1}}\langle\phi_{2,1}(r_1)\phi_{2,1}(r_2)\rangle

1870: \propto\left({\epsilon\over|r_1-r_2|}\right)^{2h_{2,1}}\,.

1871: \end{equation}

1872: Such a result, elementary in CFT, is difficult to obtain directly

1873: from SLE in which the curves are conditioned to begin and end at

1874: \em given \em points.

1875: 

1876: Note that the central charge $c$ vanishes when either locality

1877: ($\kappa=6$) or restriction ($\kappa=\frac83$) hold. These cases

1878: correspond to the continuum limit of percolation and self-avoiding

1879: walks respectively, corresponding to formal limits $Q\to1$ in the

1880: Potts model and $n\to0$ in the $O(n)$ model for which the

1881: unconditioned partition function is trivial.

1882: 

1883: \subsection{Differential Equations}

1884: In this section we discuss how the linear second order

1885: differential equations for various observables which arise from

1886: the stochastic aspect of SLE follow equivalently from the null

1887: state condition in CFT. In this context they are known as the BPZ

1888: equations\cite{BPZ}. As an example consider Schramm's formula

1889: (\ref{eqn:schrammformula}) for the probability $P$ that a point

1890: $\zeta$ lies to the right of $\gamma$, or equivalently the

1891: expectation value of the indicator function ${\cal O}(\zeta)$

1892: which is 1 if this is satisfied and zero otherwise. In SLE, this

1893: expectation value is with respect to the measure on curves which

1894: connect the point $a_0$  to infinity. In CFT, as explained above,

1895: we can only consider curves which intersect some

1896: $\epsilon$-neighbourhood on the real axis. Therefore $P$ should be

1897: written as a ratio of expectation values with respect to the CFT

1898: measure:

1899: \begin{equation}

1900: P(\zeta;a_0)=\lim_{r_2\to\infty}{\langle\phi_{2,1}(a_0){\cal

1901: O}(\zeta)\phi_{2,1}(r_2)\rangle\over

1902: \langle\phi_{2,1}(a_0)\phi_{2,1}(r_2)\rangle}\,.

1903: \end{equation}

1904: We can derive differential equations for the correlators in the

1905: numerator and denominator by inserting into each of them a factor

1906: $(1/(2\pi i)\int_\Gamma\alpha(z)T(z)dz+$ c.c., where

1907: $\alpha(z)=2/(z-a_0)$, and $\Gamma$ is a small semicircle

1908: surrounding $a_0$. This is equivalent to making the infinitesimal

1909: transformation $z\to z+2\epsilon/(z-a_0)$. As before, the

1910: c.c.~term is equivalent to extending the contour in the first term

1911: to a full circle. The effect of this insertion may be evaluated in

1912: two ways: by shrinking the contour onto $a_0$ and using the OPE

1913: between $T$ and $\phi_{2,1}$ we get

1914: \begin{equation}\label{var1}

1915: 2L_{-2}\phi_{2,1}(a_0)=(\kappa/2)L_{-1}^2\phi_{2,1}(a_0)=

1916: \partial^2_{a_0}\phi_{2,1}(a_0)\,,

1917: \end{equation}

1918: while wrapping it about $\zeta$ (in a clockwise sense) we get

1919: \begin{equation}\label{var2}

1920: -\big(2/(\zeta-a_0)\big)\partial_\zeta{\cal

1921: O}-\big(2/(\bar\zeta-a_0)\big)\partial_{\bar\zeta}{\cal O}\,.

1922: \end{equation}

1923: The effect on $\phi_{2,1}(r_2)$ vanishes in the limit

1924: $r_2\to\infty$. As a result we can ignore the variation of the

1925: denominator in this case. Equating (\ref{var1}) and (\ref{var2})

1926: inside the correlation function in the numerator then leads to the

1927: differential equation (\ref{eqn:pde1}) for $P$ found in

1928: Sec.~\ref{sec:schrammformula}.

1929: 

1930: \subsubsection{Calagero-Sutherland model}\label{sec:CS}

1931: While many of the results of SLE may be re-derived in CFT with less

1932: rigour but perhaps greater simplicity, the latter contains much

1933: more information which is not immediately apparent from the SLE

1934: perspective. For example, one may consider correlation functions

1935: $\langle\phi_{1,2}(r_1)\phi_{1,2}(r_2)

1936: \ldots\phi_{2,1}(r_N)\ldots\rangle$ of multiple boundary condition

1937: changing operators with other operators either in the bulk or on

1938: the boundary. Evaluating the effect of an insertion $(1/2\pi

1939: i)\int_\Gamma T(z)dz/(z-r_j)$ where $\Gamma$ surrounds $r_j$ leads

1940: to a second order differential equation satisfied by the

1941: correlation function for \em each \em $j$.

1942: 

1943: This property is very powerful in the radial version. Consider the

1944: correlation function

1945: \begin{equation}

1946: C_\Phi(\theta_1,\ldots,\theta_N)=\langle\phi_{2,1}(\theta_1)

1947: \ldots\phi_{2,1}(\theta_N)\Phi(0)\rangle

1948: \end{equation}

1949: of $N$ $\phi_{2,1}$ operators on the boundary of the unit disc

1950: with a single bulk operator $\Phi$ at the origin. Consider the

1951: effect of inserting $(1/2 \pi i)\int_\Gamma\alpha_j(z)T(z)dz$

1952: where (cf. (\ref{eqn:radialsle}))

1953: \begin{equation}

1954: \alpha_j(z)=-z{z+e^{i\theta_j}\over z-e^{i\theta_j}}

1955: \end{equation}

1956: and $\Gamma$ surrounds the origin. Once again, this may be

1957: evaluated in two ways: by taking $\Gamma$ up to the boundary, with

1958: exception of small semicircles around the points $e^{i\theta_k}$,

1959: we get $G_jC_\Phi$, where $G_j$ is the second order differential

1960: operator

1961: \begin{equation}

1962: G_j=-\frac\kappa

1963: 2{\partial^2\over\partial\theta_j^2}+\frac{h_{2,1}}6+\frac c{12} -

1964: \sum_{k\not=j}\left(\cot{\theta_k-\theta_j\over2}

1965: {\partial\over\partial\theta_k}

1966: -{1\over2\sin^2(\theta_k-\theta_j)/2}h_{2,1}\right)\,.

1967: \end{equation}

1968: The first three terms come from evaluating the contour integral

1969: near $e^{i\theta_j}$, where $\alpha_j$ acts like

1970: $2L_{-2}-\frac16L_0-\frac{c}{12}$, (the term $\frac{c}{12}$ comes

1971: from the curvature of the boundary,) and the term with $k\not=j$

1972: from the contour near $e^{i\theta_k}$, where it acts like

1973: $\alpha_j(e^{i\theta_k})L_{-1}+{\rm

1974: Re}\,\alpha_j'(e^{i\theta_k})L_0$.

1975: 

1976: On the other hand, shrinking the contour down on the origin we see

1977: that $\alpha_j(z)=z+O(z^2)$, so that on $\Phi(0)$ it has the

1978: effect of $L_0+\overline L_0+\ldots$, where the omitted terms

1979: involve the $L_n$ and $\overline L_n$ with $n>0$. Assuming that

1980: $\Phi$ is primary, these other terms vanish, leaving simply

1981: $(L_0+\overline L_0)\Phi=x_\Phi\Phi$. Equating the two evaluations

1982: we find the differential equation

1983: \begin{equation}

1984: G_j\,C_\Phi=x_\Phi\,C_\Phi\,.

1985: \end{equation}

1986: 

1987: In general there is an $(N-1)$-dimensional space of independent

1988: differential operators $G_j$ with a common eigenfunction $C_\Phi$.

1989: (There is one fewer dimension because they all commute with the

1990: total angular momentum $\sum_j(\partial/\partial\theta_j)$.) For

1991: the case $N=2$, setting $\theta=\theta_2-\theta_1$, we recognise

1992: the differential operator in Sec.~\ref{sec:onearm}.

1993: 

1994: In general these operators are not self-adjoint and their spectrum

1995: is difficult to analyse. However, if we form the equally-weighted

1996: linear combination $G\equiv \sum_{j=1}^NG_j$, the terms with a

1997: single derivative may be written in the form $\sum_k (\partial

1998: V/\partial\theta_k)(\partial/\partial\theta_k)$ where $V$ is a

1999: potential function. In this case it is well known from the theory

2000: of the Fokker-Plank equation that $G$ is related by a similarity

2001: transformation to a self-adjoint operator. In fact\cite{JCCS} if

2002: we form $|\Psi_N|^{2/\kappa}G|\Psi_N|^{-2/\kappa}$ where $\Psi_N=

2003: \prod_{j<k}\big(e^{i\theta_j}-e^{\theta_k}\big)$ is the

2004: `free-fermion' wave function on the circle, the result is, up to

2005: calculable constants the well-known $N$-particle

2006: Calogero-Sutherland hamiltonian

2007: \begin{equation}

2008: H_N(\beta)=-\frac12\sum_{j=1}^N{\partial^2\over\partial\theta_j^2}

2009: +{\beta(\beta-2)\over16}\sum_{j<k}{1\over\sin^2(\theta_j-\theta_k)/2}\,,

2010: \end{equation}

2011: with $\beta=8/\kappa$. It follows that the scaling dimensions of

2012: bulk operators like $\Phi$ are simply related to eigenvalues

2013: $\Lambda_N$ of $H_N$ by

2014: \begin{equation}

2015: x_\Phi=(\kappa/N)\Lambda_N(8/\kappa)-(4/\kappa N)E_N^{\rm

2016: ff}+\ffrac16h_{2,1}+\ffrac1{12}c\,,

2017: \end{equation}

2018: where $E_N^{\rm ff}=\frac1{24}N(N^2-1)$. Similarly $C_\Phi$ is

2019: proportional to the corresponding eigenfunction. In fact the

2020: ground state (with conventional boundary conditions) turns out to

2021: correspond to the bulk $N$-leg operator discussed in

2022: Sec.~\ref{Nleg1}. The corresponding correlator is

2023: $|\Psi_N|^{2/\kappa}$.

2024: 

2025: 

2026: %

2027: \newpage

2028: %

2029: \section{Related ideas}\label{sec:relatedideas}

2030: \subsubsection{Multiple SLEs}\label{sec:multiplesles}

2031: We pointed out earlier that the boundary operators $\phi_{2,1}$

2032: correspond to the continuum limits of lattice curves which hit

2033: the boundary at a given point. For a single curve, these are

2034: described by SLE, and we have shown in that case how the resulting

2035: differential equations also appear in CFT. Using the $N$-particle

2036: generalisation of the CFT results of the previous section, we may

2037: now `reverse engineer' the problem and conjecture the

2038: generalisation of SLE to $N$ curves.

2039: 

2040: The expectation value of some observable $\cal O$ given that $N$

2041: curves, starting at the origin, hit the boundary at

2042: $(\theta_1,\ldots,\theta_N)$ is

2043: \begin{equation}

2044: P_{\cal O}(\theta_1,\ldots,\theta_N) = {F_{\cal

2045: O}(\theta_1,\ldots,\theta_N)\over F_{\bf

2046: 1}(\theta_1,\ldots,\theta_N)}\,,

2047: \end{equation}

2048: where $F_{\cal O}=\langle{\cal O}\,\phi_{2,1}(e^{i\theta_1})

2049: \ldots \phi_{2,1}(e^{i\theta_N})\Phi_N(0)\rangle$. This satisfies

2050: the BPZ equation

2051: \begin{equation}

2052: G_j\,F_{\cal O}=\langle(\delta_j{\cal O}\phi_{2,1}(e^{i\theta_1})

2053: \ldots \phi_{2,1}(e^{i\theta_N})\Phi_N(0)\rangle\,,

2054: \end{equation}

2055: where $\delta_j{\cal O}$ is the variation in $\cal O$ under

2056: $\alpha_j$. If we now write $F_{\cal O}=F_{\bf 1}\cdot P_{\cal O}$

2057: and use the fact that $G_jF_{\bf 1}=x_\Phi F_{\bf 1}$, we find a

2058: relatively simple differential equation for $P_{\cal O}$, since

2059: the non-derivative terms in $G_j$ cancel. There is also a

2060: complication since the second derivative gives a cross term

2061: proportional to $\big(\partial_{\theta_j}F_{\bf 1}\big)

2062: \big(\partial_{\theta_j}P_{\cal O}\big)$. However, this may be

2063: evaluated from the explicit form $F_{\bf 1}=|\Psi_N|^{2/\kappa}$.

2064: The result is

2065: \begin{equation}

2066: \left(\frac\kappa2{\partial^2\over\partial\theta_j^2} +

2067: \sum_{k\not=j}\cot{\theta_k-\theta_j\over2}

2068: \left({\partial\over\partial\theta_k}-

2069: {\partial\over\partial\theta_j}\right)\right)P_{\cal O}=

2070: \delta_jP_{\cal O}\,,

2071: \end{equation}

2072: where the right hand side comes from the variation in $\cal O$.

2073: 

2074: The left hand side may be recognised as the generator (the adjoint

2075: of the Fokker-Planck operator) for the stochastic process

2076: \begin{eqnarray}

2077: d\theta_j&=&\sqrt\kappa

2078: dB_t+\sum_{k\not=j}(\rho_k/2)\cot\big((\theta_j-\theta_k)/2\big)\,

2079: dt\,;\label{mult1}\\

2080: d\theta_k&=&\cot\big((\theta_k-\theta_j)/2\big)\,dt\,,\label{mult2}

2081: \end{eqnarray}

2082: where $\rho_k=2$. [For general values of the parameters $\rho_k$

2083: this process is known as (radial) SLE$(\kappa,\vec\rho\,)$,

2084: although this is more usually considered in the chordal version.

2085: It has been argued\cite{JCSLEkr} that this applies to the level

2086: lines of a free gaussian field with piecewise constant Dirichlet

2087: boundary conditions: the parameters $\rho_k$ are related to the

2088: size of the discontinuities at the points $e^{i\theta_k}$.

2089: SLE$(\kappa,\vec\rho\,)$ has also been used to give examples of

2090: restriction measures on curves which are not reflection

2091: symmetric\cite{SLEkrrefs}.]

2092: 

2093: We see that $e^{i\theta_j}$ undergoes Brownian motion but is also

2094: repelled by the other particles at $e^{i\theta_k}$ ($k\not=j$):

2095: these particles are themselves repelled deterministically from

2096: $e^{i\theta_j}$. The infinitesimal transformation $\alpha_j$

2097: corresponds to the radial Loewner equation

2098: \begin{equation}\label{mult3}

2099: {dg_{j,t}\over dt}=-g_{j,t}\,{g_{j,t}+e^{i\theta_{j,t}} \over

2100: g_{j,t}-e^{i\theta_{j,t}}}\,.

2101: \end{equation}

2102: 

2103: The conjectured interpretation of this is as follows: we have $N$

2104: non-intersecting curves connecting the boundary points

2105: $e^{i\theta_{k,0}}$ to the origin. The evolution of the $j$th

2106: curve in the presence of the others is given by the radial Loewner

2107: equation with, however, the driving term not being simple Brownian

2108: motion but instead the more complicated process

2109: (\ref{mult1},\ref{mult2}).

2110: 

2111: However, from the CFT point of view we may equally well consider

2112: the linear combination $\sum_jG_j$. The Loewner equation is now

2113: \begin{equation}\label{mult4}

2114: \dot g_t=-g_t\sum_{j=1}^N{g_t+e^{i\theta_{j,t}} \over

2115: g_t-e^{i\theta_{j,t}}}\,,

2116: \end{equation}

2117: where

2118: \begin{equation}\label{mult5}

2119: d\theta_j=\sqrt\kappa

2120: dB_t^j+2\sum_{k\not=j}\cot\big((\theta_j-\theta_k)/2\big)\,dt\,.

2121: \end{equation}

2122: This is known in the theory of random matrices as Dyson's Brownian

2123: motion. It describes the statistics of the eigenvalues of unitary

2124: matrices. The conjectured interpretation is now in terms of $N$

2125: random curves which are all growing in each other's mutual

2126: presence at the \em same \em mean rate (measured in Loewner time).

2127: From the point of view of SLE, it is by no means obvious that the

2128: measure on $N$ curves generated by process

2129: (\ref{mult1},\ref{mult2},\ref{mult3}) is the same as that given by

2130: (\ref{mult4},\ref{mult5}). However CFT suggests that, for curves

2131: which are the continuum limit of suitable lattice models, this is

2132: indeed the case.

2133: \subsection{Other variants of SLE}\label{sec:variants}

2134: So far we have discussed only chordal SLE, which describes curves

2135: connecting distinct points on the boundary of a simple connected

2136: domain, and radial SLE, in which the curve connects a boundary

2137: point to an interior point. Another simple variant is dipolar

2138: SLE\cite{dipolar}, in which the curve is constrained to start at

2139: boundary point and to end on some finite segment of the boundary

2140: not containing the point. The canonical domain is an infinitely

2141: long strip, with the curve starting a point on one edge and ending

2142: on the other edge. This set-up allows the computation of several

2143: interesting physical quantities.

2144: 

2145: The study of SLE in multiply-connected domains is very

2146: interesting. Their conformal classes are characterised by a set of

2147: moduli, which change as the curve grows. R.~Friedrich and

2148: co-workers\cite{BFriedrich} have argued that SLE in such a domain

2149: is characterised by diffusion in moduli space as well as diffusion

2150: on the boundary.

2151: 

2152: It is possible to rewrite the differential equations which arise

2153: from null state conditions in extended CFTs (for example

2154: super-conformal CFTs\cite{Ras} and WZWN models\cite{GLW}) in terms

2155: of the generators of stochastic conformal mappings which

2156: generalise that of Loewner. However, a physical interpretation in

2157: terms of the continuum limit of lattice curves appears so far to

2158: be missing.

2159: 

2160: \subsection{Other growth models}

2161: SLE is in fact just one very special, solvable, example of an

2162: approach to growth processes in two dimensions using conformal

2163: mappings which has been around for a number of years. For a recent

2164: review see \cite{BBrev}. The prototypical problem of this type is

2165: diffusion-limited aggregation (DLA). In this model of cluster

2166: formation, particles of finite radius diffuse in, one by one, from

2167: infinity until they hit the existing cluster, where they stick.

2168: The probability of sticking at a given point is proportional to

2169: the local electric field, if we imagine the cluster as being

2170: charged. The resultant highly branched structures are very similar

2171: to those observed in smoke particles, and in viscous fingering

2172: experiments where one fluid is forced into another in which it is

2173: immiscible. Hastings and Levitov\cite{HL} proposed an approach to

2174: this problem using conformal mappings. At each time $t$, the

2175: boundary of the cluster is described by the conformal mapping

2176: $f_t(z)$ which takes it to the unit disc. The cluster is grown by

2177: adding a small semicircular piece to the boundary. The way this

2178: changes $f_t$ is well known according to a theorem of Hadamard.

2179: The difficulty is that the probability of adding this piece at a

2180: given point depends on the local electric field which itself

2181: depends on $f_t'$. The equation of motion for $f_t$ is therefore

2182: more complicated than in SLE. Moreover it may be shown that almost

2183: all initially smooth boundary curves evolve towards a finite-time

2184: singularity: this is thought to be responsible for branching, but

2185: just at this point the equations must be regularised to reflect

2186: the finite size of the particles (or, in viscous fingering, the

2187: effects of finite surface tension.)

2188: 

2189: It is also possible to generate branching structures by making the

2190: driving term $a_t$ in Loewner's equation discontinuous, for

2191: example taking it to be a Levy process. Unfortunately this does

2192: not appear to describe a physically interesting model.

2193: 

2194: Finally, Hastings\cite{Hastings} has proposed two related growth

2195: models which each lead, in the continuum limit, to SLE. These are

2196: very similar to DLA, except that growth is only allowed at the

2197: tip. The first, called the arbitrary Laplacian random walk, takes

2198: place on the lattice. The tip moves to one of the neighbouring

2199: unoccupied sites $r$ with relative probability $E(r)^\eta$, where

2200: $E(r)$ is the lattice electric field, that is the potential

2201: difference between the tip and $r$, and $\eta$ is a parameter. The

2202: second growth model takes place in the continuum via iterated

2203: conformal mappings, in which pieces of length $\ell_1$ are added

2204: to the tip, but shifted to the left or right relative to the

2205: previous growth direction by a random amount $\pm\ell_2$. This

2206: model depends on the ratio $\ell_2/\ell_1$, and leads, in the

2207: continuum limit, to SLE$_\kappa$ with $\kappa=\ell_2/4\ell_1$. For

2208: the lattice model there is no universal relation between $\kappa$

2209: and $\eta$, except for $\eta=1$, which is the same as the

2210: loop-erased random walk (Sec.~\ref{sec:Onmodel}) and converges to

2211: SLE$_2$.

2212: %

2213: \begin{thebibliography}{99}

2214: %

2215: \bibitem{JCcrossing} J.~Cardy, {\sl Critical percolation in finite

2216: geometries}, J. Phys. A {\bf 25}, L201, 1992 (hep-th/9111026).

2217: %

2218: \bibitem{Lang} R.~Langlands, P.~Pouliot and Y.~Saint-Aubin,

2219: {\sl Conformal invariance and two-dimensional percolation,}

2220: Bulletin of the AMS {\bf 30}, 1, 1994 (math.MP/9401222).

2221: %

2222: \bibitem{Nienhuis} B.~Nienhuis, {\sl Coulomb gas formulation of

2223: two-dimensional phase transitions,} in {\sl Phase transitions and

2224: critical phenomena}, vol. 11, C.~Domb and J.L.~Lebowitz, eds.

2225: (Academic, 1987.)

2226: %

2227: \bibitem{Loewner} K.~L\"owner, {\sl Untersuchungen \"uber schlichte

2228: konforme Abbildungen des Einheitskreises,} I, Math. Ann. {\bf 89},

2229: 103, 1923.

2230: %

2231: \bibitem{Schramm} O.~Schramm, {\sl Scaling limits of loop-erased random

2232: walks and uniform spanning trees,} Israel J. Math. {\bf 118}, 221,

2233: 2000 (math.PR/9904022).

2234: %

2235: \bibitem{LSW} G.F.~Lawler, O.~Schramm and W.~Werner, {\sl Values of

2236: Brownian intersection exponents I: Half-plane exponents,} Acta

2237: Math. {\bf 187(2)}, 237, 2001 (math.PR/9911084); {\sl II: Plane

2238: exponents,} \em ibid.\em  {\bf 187(2)}, 275, 2001

2239: (math.PR/0003156); {\sl III: Two-sided exponents,} Ann. Inst. H.

2240: Poincar\'e Statist. {\bf 38(1)}, 109, 2002 (math.PR/0005294).

2241: %

2242: \bibitem{Smirnov} S.~Smirnov, {\sl Critical percolation in the plane:

2243: conformal invariance, Cardy's formula, scaling limits,} C. R.

2244: Acad. Sci. Paris S\'er. I Math. {\bf 333(3)}, 239, 2001; a longer

2245: version is available at http://www.math.kth.se/$\sim$stas/papers.

2246: %

2247: \bibitem{SmirnovWerner} S.~Smirnov and W.~Werner, {\sl Critical

2248: exponents for two-dimensional percolation,} Math. Res. Lett. {\bf

2249: 8}, 729, 2001; (math.PR/0109120).

2250: %

2251: \bibitem{JCsurf} J.~Cardy, {\sl Conformal Invariance and Surface Critical

2252: Behavior,} Nucl. Phys. B {\bf 240}, 514, 1984.

2253: %

2254: \bibitem{BB} M.~Bauer and D.~Bernard, {\sl Conformal field theories

2255: of stochastic Loewner evolutions}, Comm. Math. Phys. {\bf 239},

2256: 493, 2003 (hep-th/0210015); {\sl SLE$_\kappa$ growth processes and

2257: conformal field theory,} Phys. Lett. B {\bf 543}, 135, 2002

2258: (math-ph/0206028); {\sl SLE martingales and the Virasoro algebra,}

2259: Phys. Lett. B {\bf 557}, 309, 2003 (hep-th/0301064); {\sl CFTs of

2260: SLES: the radial case,} Phys. Lett. B {\bf 583}, 324, 2004

2261: (math-ph/0310032); {\sl Conformal transformations and the SLE

2262: partition function martingale,} Ann. Henri Poincar\'e {\bf 5},

2263: 289, 2004 (math-ph/0305061).

2264: %

2265: \bibitem{FriedrichWerner} R.~Friedrich and W.~Werner, {\sl Conformal

2266: fields, restriction properties, degenerate representations and

2267: SLE,} C.R. Acad. Sci. Paris, Ser. I Math. {\bf 335}, 947

2268: (math.PR/0209382); {\sl Conformal restriction, highest-weight

2269: representations and SLE,} Comm. Math. Phys. {\bf 243}, 105, 2003

2270: (math-ph/0301018).

2271: %

2272: \bibitem{Wreview} W.~Werner, {\sl Random planar curves and

2273: Schramm-Loewner evolutions,} in {\sl Ecole d'Et\'e de

2274: Probabilit\'es de Saint-Flour XXXII (2002)}, Springer Lecture

2275: Notes in Mathematics {\bf 1180}, 113, 2004 (math.PR/0303354).

2276: %

2277: \bibitem{Lreview} G.F.~Lawler, {\sl Conformally invariant processes in the

2278: plane,} ICTP Lecture Notes {\bf 17}, 305, 2004

2279: (http://www.ictp.trieste.it/$\sim$pub\_off/lectures/vol17.html);

2280: {\sl Stochastic Loewner Evolution,} to appear in {\sl Encyclopedia

2281: of Mathematical Physics}, J.-P.~Fran\c coise, G.~Naber and

2282: T.S.~Tsun, eds. (Elsevier, 2005.)

2283: (http://www.math.cornell.edu/~lawler/encyclopedia.ps); {\sl

2284: Conformally Invariant Processes in the Plane}, book, (American

2285: Math. Soc., 2005).

2286: %

2287: \bibitem{KNreview} W.~Kager and B.~Nienhuis, {\sl A guide to stochastic

2288: Loewner evolution and its applications,} J. Stat. Phys. {\bf 115},

2289: 1149, 2004 (math-ph/0312251).

2290: %

2291: \bibitem{KadGruz} I.A.~Gruzberg and L.P.~Kadanoff, {\sl The Loewner

2292: equation: maps and shapes,} J. Stat. Phys. {\bf 114}, 1183, 2004

2293: (cond-mat/0309292).

2294: %

2295: \bibitem{Dupreview} B.~Duplantier, {\sl Conformal fractal geometry and

2296: boundary quantum gravity,} math-ph/0303034.

2297: %

2298: \bibitem{Kondev} J.~Kondev, {\sl Liouville theory of fluctuating

2299: loops,}

2300: Phys. Rev. Lett. {\bf 78}, 4320, 1997 (cond-mat/9703113).

2301: %

2302: \bibitem{KadKager} W.~Kager, B.~Nienhuis and L.P.~Kadanoff, {\sl Exact

2303: solutions for Loewner evolutions,} J. Stat. Phys. {\bf 115}, 805,

2304: 2004 (math-ph/0309006).

2305: %

2306: \bibitem{RS} S.~Rohde and O.~Schramm, {\sl Basic properties of

2307: SLE,}

2308: Ann. Math., to appear (math.PR/0106036).

2309: %

2310: \bibitem{Dupduality} B.~Duplantier, {\sl Conformally invariant fractals

2311: and potential theory,} Phys. Rev. Lett. {\bf 84}, 1363, 2000

2312: (cond-mat/9908314).

2313: %

2314: \bibitem{Beffaraduality} V.~Beffara, {\sl Hausdorff dimensions for

2315: SLE$_6$,} Ann. Probab. {\bf 32}, 2606, 2004 (math.PR/0204208).

2316: %

2317: \bibitem{LSWSAW} G.F.~Lawler, O.~Schramm and W.~Werner, {\sl On the

2318: scaling limit of the planar self-avoiding walk,} to appear in {\sl

2319: Fractal geometry and applications, a jubilee of Benoit

2320: Mandelbrot}, AMS Proc. Symp. Pure Math. (math.PR/02044277).

2321: %

2322: \bibitem{SSHE} O.~Schramm and S.~Sheffield, {\sl The harmonic explorer

2323: and its convergence to SLE$_4$} (math.PR/0310210).

2324: %

2325: \bibitem{LSWUST} G.F.~Lawler, O.~Schramm and W.~Werner, {\sl Conformal

2326: invariance of planar loop-erased random walks and uniform spanning

2327: trees,} Ann. Probab. {\bf 32}, 939, 2004 (math.PR/0112234).

2328: %

2329: \bibitem{Schrammformula} O.~Schramm, {\sl A percolation formula,}

2330: Electronic Comm. Probab. {\bf 8}, paper no. 12, 2001

2331: (math.PR/0107096).

2332: %

2333: \bibitem{Langlandsetal} R.P.~Langlands, C.~Pichet, Ph.~Pouliot and

2334: Y.~Saint-Aubin, {\sl On the universality of crossing probabilities

2335: in two-dimensional percolation,} J. Stat. Phys. {\bf 67}, 553,

2336: 1992.

2337: %

2338: \bibitem{Watts} G.~Watts, {\sl A crossing probability for percolation in two dimensions},

2339: J. Phys. A {\bf 29}, L363, 1996 (cond-mat/9603167).

2340: %

2341: \bibitem{Dubedat} J.~Dub\'edat, {\sl Excursion decompositions for SLE$_6$ and

2342: Watts' crossing formula} (math.PR/0405074).

2343: %

2344: \bibitem{JCmean} J.~Cardy, {\sl Linking numbers for self-avoiding loops and percolation:

2345: application to the spin quantum Hall transition,} Phys. Rev. Lett.

2346: {\bf 84}, 3507, 2000 (cond-mat/9911457).

2347: %

2348: \bibitem{JCannulus} J.~Cardy, {\sl Crossing Formulae for Critical Percolation in an

2349: Annulus,} J. Phys. A {\bf 35}, L565, 2002 (math-ph/0208019).

2350: %

2351: \bibitem{Beffara} V.~Beffara, {\sl The dimension of the SLE

2352: curves} (math.PR/0211322).

2353: %

2354: \bibitem{LSW1arm} G.F.~Lawler, O.~Schramm and W.~Werner, {\sl One-arm

2355: exponent for critical 2D percolation,} Electronic J. Probab. {\bf

2356: 7}, paper no. 2, 2002 (math.PR/0108211).

2357: %

2358: \bibitem{BPZ} A.A.~Belavin, A.M.~Polyakov and A. B. Zamolodchikov,

2359: J. Stat. Phys. {\bf 34}, 763, 1984.

2360: %

2361: \bibitem{JCBCFT} J.~Cardy, {\sl Boundary Conformal Field Theory,} to appear in

2362: {\sl Encyclopedia of Mathematical Physics}, J.-P.~Fran\c coise,

2363: G.~Naber and T.S.~Tsun, eds. (Elsevier, 2005) (hep-th/0411189).

2364: %

2365: \bibitem{YellowBook} P.~Di Francesco, P.~Mathieu and D.~Senechal,

2366: {\sl Conformal field theory} (Springer, 1997).

2367: %

2368: \bibitem{JCCS} J.~Cardy, {\sl Calogero-Sutherland Model and Bulk-Boundary Correlations in

2369: Conformal Field Theory,} Phys. Lett. B {\bf 582}, 121, 2004

2370: (hep-th/0310291).

2371: %

2372: \bibitem{JCSLEkr} J.~Cardy, {\sl SLE$(\kappa,\rho)$ and Conformal Field

2373: Theory} (math-ph/0412033).

2374: %

2375: \bibitem{SLEkrrefs} J.~Dub\'edat, {\sl SLE$(\kappa,\rho)$ martingales and

2376: duality}, Ann. Probab. {\bf 33}, 223, 2005 (math.PR/0303128);

2377: W.~Werner, {\sl Girsanov's transformation for SLE$(\kappa,\rho)$

2378: processes, intersection exponents and hiding exponents,} Ann. Fac.

2379: Sci. Toulouse (6) {\bf 13}, 121, 2004 (math.PR/0302115).

2380: %

2381: \bibitem{dipolar} M.~Bauer, D.~Bernard and J.~Houdayer {\sl Dipolar SLEs}

2382: (math-ph/0411038).

2383: %

2384: \bibitem{BFriedrich} R.~Friedrich and J.~Kalkkinen, {\sl On conformal

2385: field theory and stochastic Loewner evolution,} Nucl. Phys. B {\bf

2386: 687}, 279, 2004 (hep-th/03087023); R.~Friedrich, {\sl On

2387: connections of conformal field theory and stochastic Loewner

2388: evolution} (math-ph/0410029); R.O.~Bauer and R.~Friedrich, {\sl On

2389: chordal and bilateral SLE in multiply connected domains}

2390: (math.PR/0503178).

2391: %

2392: \bibitem{Ras} J.~Nagi and J.~Rasmussen, {\sl On stochastic evolutions

2393: and super-conformal field theory,} Nucl. Phys. B {\bf 704}, 475,

2394: 2005 (math-ph/0407049).

2395: %

2396: \bibitem{GLW} J.~Rasmussen, {\sl On SU$(2)$

2397: Wess-Zumino-Witten models and stochastic evolutions}

2398: (hep-th/0409026); E.~Bettelheim, I.~Gruzberg, A.W.W.~Ludwig and

2399: P.~Wiegmann, {\sl Stochastic Loewner evolution for conformal field

2400: theories with Lie group symmetries} (hep-th/0503013).

2401: %

2402: \bibitem{BBrev} M.~Bauer and D.~Bernard, {\sl Loewner Chains}

2403: (cond-mat/0412372).

2404: %

2405: \bibitem{HL} M. Hastings and L.S.~Levitov, {\sl Laplacian growth as one-dimensional

2406: turbulence,} Physica D {\bf 116}, 244, 1998 (cond-mat/9607021).

2407: %

2408: \bibitem{Hastings} M.~Hastings, {\sl Exact multi-fractal spectra for

2409: arbitrary laplacian random walks,} Phys. Rev. Lett. {\bf 88},

2410: 055506, 2002 (cond-mat/0109304).

2411: %

2412: \end{thebibliography}

2413: \end{document}

2414: