math0007042/s2.tex
1: 
2: \documentclass{article}
3: 
4: \usepackage{amssymb, graphics}
5: 
6: %%\nofiles
7: %%\writetoctrue
8: 
9: %%\input defs
10: \hfuzz 20pt
11: \newcommand{\R}{\mathbb{R}}
12: \newcommand{\C}{\mathbb{C}}
13: \newcommand{\Z}{\mathbb{Z}}
14: \newcommand{\N}{\mathbb{N}}
15: \def \T {{\mathcal T}}
16: \def \P {{P}}
17: \def \V {{\mathcal V}}
18: \def \L {{L}}
19: \def \H {{H}}
20: \def \a {{\zeta}}
21: \def \eps {{\epsilon}}
22: 
23: \newtheorem{theorem}{Theorem}
24: \newtheorem{lemma}[theorem]{Lemma}
25: \newtheorem{corollary}[theorem]{Corollary}
26: \newtheorem{proposition}[theorem]{Proposition}
27: \newtheorem{claim}[theorem]{Claim}
28: \newtheorem{remark}[theorem]{Remark}
29: \newtheorem{example}[theorem]{Example}
30: \newtheorem{conjecture}[theorem]{Conjecture}
31: \newtheorem{definition}[theorem]{Definition}
32: \newtheorem{problem}[theorem]{Problem}
33: \newtheorem{prediction}[theorem]{Prediction}
34: 
35: \newenvironment{Proof}{\removelastskip\par\medskip
36: \noindent{\em Proof.} \rm}{\penalty-20\null\hfill$\square$\par\medbreak}
37: 
38: \newenvironment{Proofx}{\removelastskip\par\medskip
39: \noindent{\em Proof.} \rm}{\par}
40: 
41: \begin{document}
42: \setcounter{page}{1}
43: 
44: \title
45: {Critical exponents, conformal invariance
46: \\
47: and planar Brownian motion}
48: \author{Wendelin Werner}
49: 
50: \date {5 July 2000}
51: 
52: \maketitle
53: 
54: 
55: \begin{abstract}
56: In this review paper, we first discuss
57:  some open problems related to 
58: two-dimensional self-avoiding paths and critical 
59: percolation. We then review some closely related 
60: results (joint work with Greg Lawler and Oded Schramm)
61: on critical exponents for two-dimensional 
62: simple random walks, Brownian motions and other
63: conformally invariant random objects. 
64: \end{abstract}
65: 
66: 
67: \section {Introduction}
68: 
69: The conjecture that the scaling limits of 
70: many two-dimensional systems in statistical physics 
71: exhibit conformally invariant behaviour at criticality has
72: led to striking predictions by theoretical
73: physicists  concerning, for instance, 
74: the values of exponents that
75: describe the behaviour of certian quantities
76:  near (or at) the critical temperature.
77: Some of these predictions can be reformulated in 
78: elementary terms (see for instance the
79:   conjectures for the number of
80: self-avoiding walks of length $n$ on a planar lattice).
81: 
82: From a mathematical perspective,
83:  even if the statement of the conjectures
84: are clear, the understanding of these predictions and of the 
85: non-rigorous techniques 
86: (renormalisation group, conformal field theory,
87: quantum gravity, the link with
88: highest-weight representation of some
89: infinite-dimensional Lie algebras, see 
90: e.g.  \cite {ID,Cabook}) used by physicists
91:  has been limited.
92: Our aim in the present review paper is to present 
93: some results derived in joint work with Greg Lawler and
94: Oded Schramm \cite {LW1,LW2,LSW1, LSW2, LSW3,LSWa}
95: that  proves some of these conjectures, and
96: improves substantially our understanding of others.
97: The systems that we will focus on (self-avoiding walks,
98: critical percolation, simple random walks) correspond in the 
99: language of conformal field theory
100: to zero central charge.
101:  
102: 
103: We structure this paper as follows:
104: In order to put our results into
105: perspective, we start by very briefly describing
106:  two models (self-avoiding
107: paths and critical percolation) and some of the 
108: conjectures that theoretical physicists have produced 
109: and that are, at present, open mathematical  problems. 
110: Then, we state theorems derived in joint work with 
111: Greg Lawler and Oded Schramm \cite {LSW1,LSW2,LSWa}
112: concerning
113: critical exponents for simple random walks and planar
114: Brownian motion (these had been also predicted 
115: by theoretical physics). We then show how all these 
116: problems are mathematically related, and,  in particular, why 
117: the geometry of critical percolation
118: in its scaling limit,
119: should be closely related to the geometry of a planar Brownian 
120: path via a new increasing set-valued process introduced 
121: by Schramm in \cite {S}.
122:  
123: 
124: \section {Review of some prediction of theoretical physics}
125: 
126: \subsection {Predictions for self-avoiding walks}
127: 
128: We first very briefly  describe some  
129:  predictions of theoretical physics
130: concerning 
131: self-avoiding paths in a planar lattice.
132: For a more detailed mathematical  account on this subject, see 
133: for instance \cite {MS}.
134:  
135: Consider the square lattice $\Z^2$ and define the 
136: set  $\Omega_n$ of nearest-neighbour paths  of length
137: $n$ started at the origin that are self-avoiding. 
138: In other words, $\Omega_n$ is the set of injective 
139: functions $\{ 0 , \ldots, n \} \to \Z^2$, such 
140: that $w(0) = (0,0)=0$ and
141: $| w(1) - w(0) | = \cdots = | w(n) - w(n-1) | = 1 $.
142:  
143: The first problem is to understand the asymptotic behaviour of 
144: the number $a_n := \# \Omega_n$
145:  of such self-avoiding paths when $n \to \infty$.
146: A first trivial observation is that for all $n, 
147: m \ge 1$, $a_{n+m} \le a_n a_m$ because 
148: the first $n$ steps and the last $m$ steps
149:  of a $n+m$ long self-avoiding path are self-avoiding paths
150: of length $n$ and $m$ respectively.
151: Furthermore,  $a_n \ge 2^n$ because if the path goes
152: only upward or to the right, then it is self-avoiding. 
153: This leads immediately to the existence of a constant 
154: $\mu \in [2,3) $ (called the connectivity
155: constant of the lattice $\Z^2$)
156: such that 
157: $$
158: \mu:=  \inf_{n \ge 1} (a_n)^{1/n} =
159: \lim_{n \to \infty} (a_n)^{1/n}
160: .$$
161: 
162: Note that if one counts the number $a_n'$ 
163: of self-avoiding paths 
164: of length $n$ on a triangular lattice, the   
165: same argument shows that 
166: $(a_n')^{1/n}$ converges when $n \to \infty$ to 
167: some limit $\mu' \ge 3$.
168: The connectivity constant is lattice-dependent.
169: 
170: One can also look at
171: other regular planar lattices, such as the honeycombe lattice.
172: We will say that a property is  `lattice-independent' if it 
173: it holds for all these three `regular' lattices
174: (square lattice, triangular lattice, honeycombe lattice).
175:  One possible way to 
176: describe a larger class
177: of `regular'  lattices for which  our `lattice-independent'
178: properties should hold, could   
179: (but we do not
180: want to discuss this issue in
181: detail here) be that they 
182: are transitive (i.e. for each pair of points, there exists 
183: a euclidean isometry that maps the lattice onto itself and
184: one of the two points onto the other) and that rescaled 
185: simple random walk on this lattice
186: converges to
187: planar Brownian motion. For instance, the lattice
188: $\Z \times 2\Z$ is not allowed.
189:  
190: When $L$ denotes such  a 
191: planar lattice, we denote by $a_{n,L}$ the number 
192: of self-avoiding paths of length $n$ in the lattice
193:  starting from 
194: a fixed point, and by $\mu_{L}
195: := \inf_{n \ge 1} (a_{n,L})^{1/n}$ its connectivity 
196: constant.
197:  
198: 
199: A first striking prediction from theoretical 
200: physics is the following:
201: 
202: \begin {prediction}[Nienhuis \cite {N1}]
203: For any regular planar lattice ${\L}$,
204: when $n \to \infty$,
205: $$
206: a_n (L)
207:  =  (\mu_{L})^n n^{11/32 + o(1)}.
208: $$
209: \end {prediction}
210:   
211: The first important feature is the rational exponent 
212: $11/32$. The second one is that this result 
213: does not depend on the lattice 
214: i.e., the first-order term 
215: is lattice-dependent while the second is ``universal''.
216: 
217: The following statement is almost equivalent to the 
218: previous prediction. Suppose that we define under  the
219: same probability  $P_n$ two independent 
220: self-avoiding paths $w$ and $w'$ of length  
221: $n$ on the lattice ${L}$
222: (the law of $w$ and $w'$ is  the uniform probability on $\Omega_n$).
223: 
224: \begin {prediction}[Intersection exponents version]
225: When $n \to \infty$,
226: $$
227: P_n [ w  \{ 1, 2, \ldots , n\} \cap w'\{ 0,1, \ldots ,n \}
228: = \emptyset ] 
229: = n^{-11/32 + o(1)}
230: . $$
231: \end {prediction}
232:   
233: Indeed,  $w  \{ 1, 2, \ldots , n\}$ and
234: $w'\{ 0,1, \ldots ,n \}$
235: are disjoint if and only if the concatenation 
236: of the  two paths $w$ and $w'$ is a   self-avoiding 
237: path of length $2n$ so that 
238: the non-intersection probability is exactly
239: $a_{2n}/ (a_n)^2$.
240:  
241: A second question concerns the typical behaviour of 
242: a long self-avoiding path, chosen uniformly in $\Omega_n$ when $n$ 
243: is large.
244: Let $d(w)$ denote the diameter of 
245: $w$. 
246: Theoretical physics 
247: predicts that the typical diameter is of
248: order $n^{3/4}$. This had 
249: already been predicted using a different
250: (`very non-rigorous') 
251: argument by Flory \cite {Fl}
252: in the late 40's.
253: A formal way to describe this prediction is
254: the following:
255: 
256: \begin {prediction}[Nienhuis \cite {N1}]
257: For all $\epsilon >0$
258: and all regular lattices, 
259: when $n \to \infty$,
260: $$
261: P_n \left[ d (w) \in  [n^{3/4 - \epsilon},
262: n^{3/4 + \epsilon} ]\right] \to 1 
263: .$$
264: \end {prediction}
265: 
266: 
267: One of the underlying beliefs that lead to these conjectures 
268: is that the measure on long  self-avoiding paths,
269: suitably rescaled, converges when the length goes to infinity,
270:  towards
271: a measure on continuous curves,
272: that posesses some
273: invariance properties under conformal transformations.
274: The counterparts of the previous predictions
275: in terms of this limitting measure then 
276: go as follows:
277: Take two independent paths defined under the 
278: limitting measure, started at distance $\epsilon$
279: from each other. Then, the probability that the 
280: two paths are disjoint decays like
281: $\epsilon^{11/24}$ when $\epsilon \to 0$.
282: For the second prediction: 
283: the Hausdorff dimension of a path defined under 
284: the limitting measure is almost surely $4/3$.
285: 
286: 
287: 
288: \subsection {Predictions for critical planar percolation}
289: 
290: We now  review some results
291: predicted by theoretical physics concerning critical planar 
292: percolation. A more detailed acount on these conjectures 
293: for mathematicians can be found for instance in \cite {LPS}.
294: See \cite {G} for a general introduction to percolation. 
295: 
296: Let $p\in (0,1)$ be fixed.
297: For each edge between
298: neighbouring points of the lattice, 
299: erase the edge with probability $1-p$ and 
300: keep it (and call the edge  open)
301: with probability $p$ independently 
302: for all edges. In other words,   
303: for each edge we toss a biased coin to decide 
304: whether it is erased or not.
305: This procedure defines a random subgraph of 
306: the square grid. It is not difficult to see that 
307: the large-scale geometry of this subgraph depends
308: a lot on the value of $p$. 
309: In particular, there exists a critical value  $p_c$ (called the 
310: critical probability), such that if $p<p_c$, 
311: there exists almost surely no unbounded connected component 
312: in the random subgraph, while if $p>p_c$,
313: there exists almost surely a unique unbounded connected component
314: of open edges.
315: It is 
316: not very difficult to see 
317: that the  value $p_c$
318: of the critical probability is lattice-dependent.
319: Kesten 
320: has shown that for $\L=\Z^2$, $p_c = 1/2$.
321: We are going to be interested in the geometry of large 
322: connected component when $p = p_c$.
323: 
324: In regular planar  lattices at $p=p_c$ it is  known that 
325: almost surely  no infinite connected component exists.
326: However, a simple duality argument shows that in the
327: square grid, at $p=p_c =1/2$, for 
328: any  $n \ge 1$, with probability $1/2$, there
329: exists a path of open edges joining (in that rectangle) 
330: the bottom and top boundaries of a fixed  $n  \times (n+1)$  rectangle.
331: This loosely speaking shows that in a big box, with 
332: large probability, there
333: exist connected components of diameter 
334: comparable to the size of the box.
335:  
336: Theoretical physics predicts that large-scale
337: properties of the geometry of critical 
338: percolation (i.e., percolation on a planar lattice at
339: its critical probability) are lattice-independent
340: (even though the value of the critical probability is
341: lattice-dependent), and, 
342: in the scaling limit, invariant under conformal transformations;
343: see e.g. \cite {Ai,LPS}.
344: Furthermore, physicists have produced  explicit 
345: formulas that describe some of its features.
346:  
347: A first prediction is the following:
348: Consider critical percolation restricted to an $n \times n$
349: square (in the square lattice, say), and choose
350: the connected component $C$ with largest diameter 
351: (among all connected components). 
352: The previous observation shows that the diameter of $C$ 
353: is of the order of magnitude
354: of $n$.
355: Define the rescaled discrete outer perimeter of $C$,
356:  $\partial_n
357: =\partial C / n$, where $\partial C$ is the boundary of the 
358: unbounded connected component of the complement of $C$ in the
359: plane.  
360: 
361: \begin {prediction}[Cluster boundaries \cite {DS,Ca,ADA}]
362: \label {cb}
363: The law of $\partial_n$ converges when $n \to \infty$ 
364: towards a law $\mu$ on continuous paths $\partial$.
365: Moreover, $\mu$-almost surely: the Hausdorff dimension of  $\partial$ 
366: is $7/4$, the path $\partial $ is not self-avoiding, and the 
367: outer boundary of $\partial$ has Hausdorff dimension $4/3$.
368: \end {prediction}
369: 
370: \noindent
371: A weaker version of the first part of this prediction is that
372: for all $\eps>0$,
373: when $n \to \infty$,
374: $
375: \P [ \# \partial C  \in ( n^{7/4 - \epsilon} , n^{7/4 + \epsilon } )  
376: ] \to   1
377: $.
378: 
379: A second prediction concerns the crossing probabilities 
380: of a quadrilateral. Suppose that $L>0$ and $l>0$, and
381: perform critical percolation in the rectangle
382: $[0, a_n] \times [0, b_n]$ where $a_n$ and $b_n$ 
383: are the respective integer parts of $Ln$ and $ln$.
384: Let $x(L,l)$ denote the cross-ratio between the 
385: four corners of the $L \times l$ rectangle (more precisely,
386: it is the value $x$ such that there
387: exists a conformal mapping from the rectangle onto the 
388: upper half-plane, such that the left and right-hand side 
389: of the rectangle are mapped onto the intervals
390: $(-\infty, 0]$ and $[1-x,1]$).
391: 
392:  
393: \begin {prediction}[Cardy's formula \cite {Ca2}]
394: When $n \to \infty$, the probability that there exists a 
395: path of open edges in the rectangle $[0, a_n]
396: \times [0,b_n]$ joining  the left  
397: and right-hand sides of the boundary of the
398: rectangle
399: converges  
400: to  
401: $$ F(x) =  \frac {3 \Gamma (2/3)}{\Gamma (1/3)^2} x^{1/3}
402: \null_2F_1 (1/3, 2/3, 4/3; x )$$
403: where $_2F_1$ is the usual hypergeometric function.
404: \end {prediction}
405: 
406: \noindent
407: These results are believed to be lattice-independent.
408: This second statement  has been predicted by 
409: Cardy \cite {Ca1,Ca2}, using and generalising 
410: conformal field 
411: theory considerations and ideas introduced in 
412: \cite {BPZ1,BPZ2}. 
413: Note that this prediction is of a different nature than the
414: previous ones. It gives an exact formula for
415: an event in the scaling limit rather than just an exponent.
416: Assuming conformal invariance, this prediction 
417: can be reformulated in a half-plane as follows:
418: 
419: \begin {prediction}[Cardy's formula in a half-plane]
420: For all $a, b >0$, the probability that there exists a  
421: crossing (a path of open edges) joining
422: $(- \infty , -an]$ to $[0, bn]$ in the upper half-plane
423: converges when $n \to \infty$ towards 
424: $F(b/ (a+b))$. 
425:  \end {prediction} 
426: 
427: \noindent
428: Carleson was the first to note
429: that Cardy's formula takes on a very simple form in an equilateral
430: triangle.  Suppose $A,B,C$ are the vertices of an equilaterial 
431: triangle,
432: say $A=0, B=e^{2i\pi /3}, C = e^{i\pi /3}$.
433: 
434: 
435: \begin {prediction}[Cardy's formula in an equilateral triangle]
436: In the scaling limit (performing 
437: critical percolation on the grid $\eps \Z^2$, say, and 
438: letting $\eps \to 0$), 
439:  the law of the left-most point on  
440:  $[B,C]$
441:  that is connected
442: (in the triangle)
443:  to  $[A,C]$ is the uniform distribution on $[B,C]$. 
444: \end {prediction} 
445: 
446: 
447: 
448: \section {Brownian exponents}
449: 
450: We now come to the core of the 
451: present paper and state some of the results derived 
452: in the series of papers \cite {LSW1,LSW2,LSW3,LSWa}
453: concerning critical exponents for planar Brownian
454: motions and simple random walks.
455: These are mathematical results (as opposed to the predictions
456: reviewed in the previous section) that had been  
457: predicted some 15 or 20 
458: years ago. The  proofs of these
459: theorems (that we shall briefly outline in the coming sections)
460: do 
461: use conformal invariance, complex analysis and univalent functions.
462: 
463: 
464: \subsection {Intersection exponents}
465: 
466: Suppose that $(S_n, n \ge 0)$ and 
467: $(S_n', n \ge 0)$ are two independent 
468: simple random walks on the lattice $\Z^2$
469: that are both started from the origin.
470: We are interested in the asymptotic 
471: behaviour (when $n \to \infty$) 
472: of the  probability that the traces of $S$ and $S'$ 
473: are disjoint.
474: 
475: \begin {theorem}[\cite {LSW2}]
476: \label {5/8}
477: When $n \to \infty$,
478: $$
479: \P [ S  \{ 1, 2, \ldots, n \} \cap S' \{0, 1, \ldots, 
480: n \}  = \emptyset ] 
481: =
482: n^{-5/8 + o(1) } 
483: .$$
484: \end {theorem}
485:  
486: This result had been predicted by Duplantier-Kwon \cite {DK}, see 
487: also \cite {Dqg,Dcm} for another non-rigorous derivation based
488: on quantum gravity ideas and  predictions by 
489: Khnizhnik, Polyakov and Zamolodchikov.
490:  
491: Note that, as opposed to self-avoiding walks and
492: percolation cluster boundaries, the scaling limit of 
493: planar simple random walk is well-understood mathematically:
494: It is planar Brownian motion
495: (this is lattice-independent) and it is invariant 
496: under conformal transformations (modulo time-change).
497: This is what makes it possible to prove Theorem \ref {5/8}
498: as opposed to the analogous prediction for self-avoiding walks.
499: The scaling limit analog of Theorem \ref {5/8}
500: is the following:
501: \begin {theorem}{\cite {LSW2}}
502: \label {5/8BM} 
503: Let $B$ and $B'$ denote two independent planar
504: Brownian motions started at distance 1 from each other.
505: Then,
506: when $ t \to \infty$,
507: $$
508: \P [ B[0,t] \cap B' [0,t] = \emptyset ] 
509: = t^{-5/8 + o(1)} 
510: .$$
511: \end {theorem}
512:  
513: In \cite {LSW1,LSW2,LSW3,LSWa}, analogous results concerning
514: non-intersection exponents between more than two 
515: Brownian motions (or simple random walks), in the 
516: plane or in the half-plane, are derived.
517: 
518: In fact, Theorem \ref {5/8} is a consequence of Theorem \ref {5/8BM}
519: via an invariance principle argument (see \cite {LP} and
520: the references therein for the connection between the two 
521: results).
522: 
523:  
524: \subsection {Mandelbrot's conjecture}
525:  
526: Let $(B_t, t \ge 0)$ denote a planar Brownian motion.
527: Define the hull of $B[0,1]$ as the complement 
528: of the unbounded connected component of $\C \setminus 
529: B[0,1]$ and define the 
530: the outer frontier of $B[0,1]$ as the boundary the 
531: hull of $B[0,1]$.
532: 
533: \begin {theorem}[\cite {LSW1,LSW2,LSWa}]
534: \label {mandelbrot}
535: Almost surely, the Hausdorff dimension of the 
536: outer boundary of $B[0,1]$ is $4/3$.
537: \end {theorem}
538: 
539: This result had been conjectured by Mandelbrot \cite {M}
540: based
541: on simulations and the  analogy with 
542: the conjectures for self-avoiding walks.
543: See also \cite {Dcm} for a physics approach based on quantum gravity.
544: This theorem is in fact a consequence (using  
545: results  derived by Lawler in \cite {Lhaus})
546: of the determination of 
547: the following critical exponent (called disconnection 
548: exponent):
549: 
550: \begin {theorem}[\cite {LSW1,LSW2,LSWa}]
551: If $B$ and $B'$ are two independent planar Brownian motions
552: started from $0$, 
553: then, when $t \to \infty$,
554: $$
555: P [ B[0,t] \cup B' [0,t]
556: \hbox { does not disconnect } 1 \hbox { from } \infty ]
557: = t^{-1/3 + o(1)}
558: .$$
559: \end {theorem}
560: Similarly, a consequence of Theorem \ref {5/8BM} is 
561: that the Hausdorff dimension of the set of cut 
562: points of the path $B[0,1]$ is almost 
563: surely $3/4$. Analogously, we get that the
564: set of pioneer points (i.e., points
565: $B_t$ that are on the outer boundary of $B[0,t]$)
566: in a planar Brownian path 
567: has Hausdorff dimension $7/4$, which -
568: together with Theorem \ref {mandelbrot} - is reminiscent of 
569: Prediction \ref {cb}.
570:  Also, the determination 
571: of more general exponents give the 
572: multifractal spectrum of the 
573: outer frontier of a Brownian path.
574:  See \cite {Lbuda} and the references therein
575: (papers by Lawler) for 
576: the link between critical exponents and 
577: Hausdorff dimensions. See also \cite {Bef}.
578: 
579:  \section {Universality}
580: 
581: \subsection {In the plane}
582: 
583: A first important step in the proof of  Theorems \ref {5/8BM} and
584: \ref {mandelbrot} is the  
585: observation
586: that some  conformally invariant random probability
587: measures with a special `restriction' property 
588: are identical.
589: Our presentation of universality 
590: here differs slightly from that  
591: of the preprints \cite {LW2,LSW1,LSW2,LSW3}. A more 
592: extended version of the present approach 
593: and of
594: its consequences is in preparation.
595:  
596: 
597: More precisely, let $D$ denote the open unit disc,
598: and suppose
599: that $P^0$ is a  
600: rotationally invariant probability 
601: measure defined  on the set of all 
602: simply connected compact subsets
603: $K$ of the closed unit disc, that contain the origin and 
604: such that $K \cap \partial D$ is just one single point
605: $e(K)$  (we endow this set with a well-chosen $\sigma$-field).
606: Note that the law of $e(K)$ is the uniform probability
607: $\lambda$  on 
608: $\partial D$.
609: 
610: For all $x \in D$, the probability measure $P^x$
611: is 
612: defined as the image probability measure of $P^0$
613: under a M\"obius transformation $\Phi$ 
614:  from $D$ onto $D$ 
615: such that $\Phi (0 ) = x$
616: (rotational invariance of $P^0$ shows that $P^x$
617: is independent of the actual choice of $\Phi$).
618: Similarly, for any simply connected open 
619: set $\Omega \subset \C$ (that is not identical to the whole plane) and
620: $x \in \Omega$, we define the probability $P^{x,\Omega}$
621: as the image measure of $P^0$ under a conformal 
622: transformation $\Phi$ from $D$ onto $\Omega$ with $\Phi ( 0) = x$.
623: When $\Phi$ does not extend continuously
624: to the boundary of $D$, one can view
625: $\Phi (K)$ as the union of $\Phi ( K \setminus \{ e(K)\})$
626: and the prime end $\Phi ( e(K))$.
627: 
628: 
629: \begin {definition}
630:  We say that $P^0$ is completely conformally invariant 
631: (in short: CCI) if for any simply connected $\Omega \subset D$, for any
632: $x \in D$, the measures
633: $P^{x,D}$ and $P^{x, \Omega}$ are identical when restricted to  the 
634: family of sets $\{ K  \ : \ K \cap D \subset \Omega \} $.
635: \end {definition}
636: 
637: Note that (under $P^{x, \Omega}$), $K \cap D \subset \Omega$ 
638: is equivalent to $\Phi (e(K)) \subset \partial D$.
639:  
640: 
641: 
642: 
643: \begin {theorem}
644: \label {univ}
645: There exists a unique conformally invariant probability measure.
646: It is given by the hull of a Brownian path started from 
647: $0$ and stopped at its first hitting of the unit circle.
648: \end {theorem}
649: 
650: \noindent 
651: {\bf Idea of the proof.}
652: First, note that the measure $\tilde P$ defined using 
653: the hull of a stopped  Brownian 
654: path 
655:  is indeed CCI, because of 
656: conformal invariance of planar Brownian motion and the 
657: strong Markov property.
658: Then, consider another CCI measure $P$. It is sufficient
659: to show that for all simply connected $D' \subset D$ that 
660: contains $0$,
661: \begin {equation}
662: \label {above}
663: P [ K \cap D \subset D'] = \tilde P [ K \cap D \subset D' ] 
664: \end {equation}
665: since the family of all such events $\{ K \cap D \subset D'\}$
666:  is a generating $\pi$-system
667: of the $\sigma$-field on which we define the 
668: measure.
669: Let $\Phi$ denote a conformal map from $D'$ onto 
670: $D$ with $\Phi(0) = 0 $.
671:  Then, because of the
672: CCI property,  both sides of (\ref {above}) are equal to $\lambda (
673: \Phi ( \partial D \cap \partial D' ))$ and hence equal.
674: 
675: 
676: 
677: \subsection {In the half-plane}
678: 
679: Analogous (slightly more complicated) 
680: arguments can be developped 
681: for subsets $K$ of a domain
682: that join two parts of the boundary of this domain (as opposed
683: to joining a point in the
684:  interior to the boundary, as in the previous 
685: subsection).
686:  For convenience, we  consider subsets of 
687: the equilateral 
688: triangle $A=0$, $B=e^{2i \pi/3}$, $C= e^{i \pi /3}$. 
689: 
690: 
691: We  study probability measures $P$ on the 
692: set of all simply connected 
693: compact subsets $K$ of $\overline \T$ such that
694:  $$K \cap \partial \T = [A,A_1] \cup [A, A_2] \cup \{ e(K)\}$$
695: where $A_1 = A_1(K) \in [A,B]$, $A_2= A_2 (K) \in [A,C]$ 
696: and $e(K) \in (B,C)$, and such that 
697: $ \T \setminus K$
698: consists of exactly two connected components (having
699: respectively  $[A_1, B] \cup [B, e(K)]$ and
700:  $[e(K), C] \cup [C,A_2]$ on their boundary).
701: 
702: We say that $P$ satisfies Cardy's formula if the law
703: of $e(K)$
704: is the uniform probability measure on $[B,C]$.
705: 
706: We say that $P$ satisfies Property I, if for all 
707: non-empty $(B', C') \subset [B,C]$, if $\Phi$ denotes
708: the conformal map from $\T$ onto $\T$ with $\Phi(A)=A$,
709: $\Phi(B') = B$ and $\Phi (C') = C$, the image measure of $P$
710: (restricted to $\{e(K) \in (B',C')\})$
711: under the mapping $\Phi$
712: is exactly the measure $P$ restricted to $\{A_1 \in [A,\Phi(B)]
713: \} \cap \{ A_2 \in [A, \Phi (C)]\}$.
714: 
715: Suppose that $K'$ is a  simply connected compact subset  
716: of $\overline \T$ such that 
717: $K' \cap [B,C] \not= \emptyset$, $K' \cap ([A,B] \cup [A,C] )
718: = \emptyset$, and $\T':= \T \setminus K'$ is  simply connected.
719: Let $\Phi$ denote the conformal mapping from $\T'$ onto $\T$ with 
720: $\Phi(A)=A$, $\Phi(B)=B$ and $\Phi(C)= C$.
721: We say that $P$ satisfies Property II, if for all such $K'$,
722: the image of $P$ (restricted to $\{ K \ : \ K \cap K' = \emptyset \}$)
723: under $\Phi$ is identical to the 
724: measure  $P$ restricted to the set $\{ K \ : \ K \cap \Phi
725: ( \partial \T' \setminus \partial \T) = \emptyset \}$.
726:  
727: We the say that $P$ is invariant under restriction if it 
728: satisfies Property I and Property II.
729: There are other equivalent definitions and formulations of this 
730: restriction property.
731: 
732:  
733: \begin {theorem}
734: \label {univ2}
735: There exists a unique probability measure $P$ that satisfies 
736: Cardy's formula and is invariant under restriction. It is 
737: given by the hull in $\T$ (see the precise definition 
738: below) of Brownian motion, started 
739: from $A$, reflected with oblique angle $\pi/3$ (pointing `away'
740: from $A$) on [A,B] and $[A,C]$, and stopped
741: when reaching $[B,C]$.
742: \end {theorem}
743: 
744: 
745: \begin{figure}[h]
746: \centering
747: %\vspace {-2cm}
748: \resizebox{0.5 \textwidth}{!} {\includegraphics {tri.eps}}
749: %\vspace {-1cm}
750: \caption{
751: The $\T$-hull of reflected Brownian motion
752: }
753: \end{figure}
754: 
755: 
756: Here and in the sequel, the hull
757:  in the upper half-plane 
758: (or $H$-hull) of a compact set $K \subset 
759: \overline H$
760: with $0 \in K$ is the complement of the 
761: unbounded connected component of $\overline H \setminus K$.
762: The hulll in $\T$, or $\T$-hull,  of a compact subset 
763: of $\overline \T$ that 
764: contains $A$, is the complement of the union 
765: of the connected components
766: of $\overline \T \setminus K$ that have $B$ or $C$ on their 
767: boundaries.
768: 
769: \medbreak
770: \noindent
771: {\bf Idea of the proof.}
772: Uniqueness follows easily from a similar 
773: argument as in Theorem \ref {univ}, by identifying 
774: $P ( K \cap \T \subset  {\T'})$ for a certain class of simply 
775: connected subsets $\T'$ of $\T$.  
776: For existence, one needs to verify that 
777: such a reflected Brownian motion satisfies 
778: Cardy's formula, and the restriction 
779: property follows from conformal invariance and the 
780: strong Markov property for such a reflected Brownian motion.
781: 
782: \medbreak
783: 
784: Let us briefly indicate how this 
785: reflected Brownian motion is defined (see for instance \cite {VW}
786: for details).
787: We first define it in the 
788: upper half-plane.
789: Define for any $x \in \R$, 
790: the vector $u(x) = \exp (i \pi /3)$ if $x \ge0$
791: and $u(x) = \exp (2 i \pi /3)$ if $x <0$.
792: Suppose that $B(t)$ is an ordinary planar Brownian path 
793: started from 
794: $0$. Then, there exists a unique pair $(Z_t, \ell_t)$ 
795: of continuous processes such that $Z_t$ takes its values
796: in $\overline \H$, $\ell_t$ is a non-decreasing real-valued
797: function with $\ell_0=0$
798:  that increases only when $Z_t \in \R$, and
799: $$
800: Z_t = B_t + \int_0^t u( Z_s) d\ell_s.
801: $$
802: The process $(Z_t, t \ge 0)$ is called the reflected 
803: Brownian motion in $H$ with reflection vector field $u(\cdot)$.
804: At each time $t$, we define $V_t$ as the $H$-hull
805: of $Z[0,t]$.
806: In order to define the $\T$-hull of reflected Brownian motion 
807: in the triangle that is refered to in the Theorem, consider 
808: for instance a conformal  mapping $\Phi$ from $\H$
809: onto the triangle, such that $\Phi(0)= A$,
810: $\Phi(\infty)= B$, $\Phi (M) =C$ 
811: for some real $M>0$. Then the $\T$-hull is $\Phi( V_T)$ 
812: where $T$ is the first time at which $Z$ hits $[M, \infty]$.
813: 
814: 
815: 
816: \section {Schramm's processes}
817: 
818: \subsection {In the plane}
819: 
820: It is possible to contruct a CCI 
821: measure $P$ using Loewner's differential equation 
822: (that encodes a certain class of growing families 
823: of compact sets) driven by a Brownian motion.
824: See, for instance, \cite {Dur} for a general introduction to 
825: Loewner's equation.
826: 
827: For any simply connected compact $K$ such that the
828: complement of $K$ in the plane is conformally 
829: equivalent to the complement of a disc,
830: Riemann's 
831: mapping theorem shows that there exists a unique $\alpha_K \in \R$
832: and a unique conformal
833: map $\hat f_K$
834: that maps the complement of $K$ onto the complement of 
835: the unit disk in such a way that 
836: $$ \hat f_K (z) =  z e^{-\alpha_K}  +  O(1)  $$
837: when  $z \to \infty$. 
838: 
839: Suppose now that $(\a(t), t \in \R)$ is a continuous function 
840: taking values on the unit circle.
841: For any $z \in \C$, define  $f_t (z)$ as the solution 
842: of the ordinary differential equation 
843: $$
844: \partial_t f_t (z) = -
845: f_t (z) \frac { f_t (z) + \a(t) }{f_t (z) - \a(t)}
846: $$
847: such that
848: $$
849: \lim_{t \to - \infty}  e^{-t} f_{t} (z) =   z 
850: $$
851: For any fixed $z \not= 0$, the mapping $t \mapsto f_t (z)$ is well-defined
852: up to a possibly infinite  `explosion time'
853: $T_z$, at which $f_t(z)$ hits
854: the singularity $\a(t)$ (and we put $T_0 = - \infty$).
855: Simple considerations show that for any time $t \in \R$, 
856: $\alpha_{K_t} = t$, and $f_t = \hat f_{K_t}$, where 
857: $$ K_t = \{ z \in \C \ : \ T_z \le t \}$$
858: i.e., $f_t$ is the conformal mapping from the 
859: complement of $K_t$ onto the complement of the unit disc
860: such that $f_t (z) = z e^{-t} + o(z)$ when $z \to \infty$.
861: 
862: We now take $\a(t) =  \exp ( i \sqrt {\kappa} W_{t} ), t \in \R$,
863: where $\kappa>0$ is a fixed constant and $(W_t, t \in \R)$
864: denotes a one-dimensional Brownian motion such that the law
865: of $\sqrt {\kappa} W(0)$ is the uniform distribution on $[0, 2 \pi]$.
866: We call $(K_t, t \in \R)$ Schramm's 
867: radial process with parameter $\kappa$
868: (in \cite {S,LSW2}, it is refered to as $SLE_\kappa$
869: for Stochastic Loewner Evolution process).
870:  Then, let $T$ denote the first time at which $K_t$
871: intersects the unit circle. 
872: The set $K_T$ is  a (random) 
873: simply connected compact set 
874: that intersects the unit circle at just one point.
875:  
876: \begin {theorem}{\cite {LSW1,LSW2}}
877: \label {SLE6}
878: If $\kappa=6$, the law of $K_T$ is a CCI probability 
879: measure.
880: \end {theorem}
881: 
882: In fact, a more general result (complete conformal invariance of $SLE_6$
883: as a process) also holds, see \cite {LSW1,LSW2}.
884: This theorem may seem quite surprising. Note that it 
885: fails for all other values of $\kappa$.
886:  The idea of the proof is to 
887: prove  `invariance'
888: of the law under infinitesimal deformations of $D$.
889: 
890: A direct consequence of Theorems \ref {SLE6} and \ref {univ} is that 
891: the law of $K_T$ is identical to the law of the hull of
892: a stopped Brownian path. This is a rather surprising result
893: since the two processes (hulls
894: of planar Brownian motion and $SLE_6$) are 
895: a priori very different.
896: For instance, the joint distribution at the first hitting times
897:  of
898: the circles of radius 1 and 2 are not the same for $SLE_6$ and 
899: for the hull of a planar Brownian path.
900:  
901: 
902:  
903: \subsection {In the half-plane}
904: 
905: Similarly, one can construct natural processes 
906: of compact subsets of the closed  upper-half plane 
907: $\overline \H = \{ z \ : \ \Im (z) \ge 0 \}$ that are
908: `growing
909: from the boundary'.
910: 
911: For any simply connected compact subset $K$
912: of $\overline H$ such that $0 \in K$ and $H \setminus K$ is 
913: simply connected, there exists a unique number
914: $\beta_K \ge 0$ and  a unique conformal map $\hat g_K$
915: from $H \setminus K$   
916: onto $H$ such that $\hat g_K (\infty) = \infty$ and 
917: $$
918: \hat g_K (z) = z +  \frac {\beta_K }{ z}  + o \left(
919: \frac {1}{z} \right)
920: $$
921: when $z \to \infty$.
922: $\beta_K$ which is   increasing in $K$, is 
923: a half-space analog of capacity.
924: 
925: Suppose now that $(a(t), t \ge 0)$ is a continuous
926: real-valued function, and define for all $z \in \overline \H$,
927: the solution $g_t (z)$ to the ordinary differential 
928: equation
929: $$
930: \partial_t g_t (z) = 
931: \frac { 2 }{ g_t (z) - a(t) }
932: $$
933: with $g_0 (z) = z$. This equation is well-defined up
934: to the (possibly infinite) time 
935: $
936: T_z$
937: at which $g_t(z)$ hits $a(t)$. Then, define
938: $$
939: K_t  := \{ z \in \overline \H \ : \ T_z \le t \}.
940: $$
941: $(K_t, t \ge 0)$ is an increasing family of 
942: compact sets, and it is not difficult
943: to see that for each time  $t\ge 0$, $\beta_{K_t}
944: = 2t$ and $g_t 
945: = \hat g_{K_t}$ i.e., $g_t$ is  
946: the unique conformal mapping from $\H \setminus K_t$
947: onto $\H$ such that $g_t ( \infty) = \infty$,
948: and $g_t(z) = z +  2t / z + o(1/z) $ when $z \to \infty$.
949: 
950: When $a(t) = \sqrt {\kappa} W_t$, where $W$ is one-dimensional
951: Brownian motion with $W(0)=0$, we get a random 
952: increasing family $(K_t, t \ge 0)$ that we call
953: Schramm's chordal process with parameter $\kappa$ (it is refered
954: to as chordal $SLE_\kappa$ 
955: in \cite {S,LSW1,LSW2,LSW3}).
956: 
957: Note that if we put $Z_t (z) = 
958: W_t - \kappa^{-1/2} g_t(z)$, then 
959: $$
960: Z_t (z) = W_t + \int_0^t \frac {2}{\kappa Z_s(z)} ds 
961: $$
962: so that $Z$ can be interpreted as a
963: (translation of a) complex  
964: Bessel flow of dimension $1+ (4/ \kappa)$.
965: If $\kappa<4$, it is easy to see, 
966: by comparison with a Bessel process,  
967: that  almost surely $T_z <\infty$ for all $z \in \overline \H$,
968: in other words, $\cup_{t \ge 0 } K_t = \overline H$.
969: 
970: The scaling properties of
971: Brownian motion easily show that it is possible to define 
972: (modulo increasing time-reparametrization) the law
973: of an increasing process of hulls
974: in any simply connected set. More precisely, if 
975: $\Phi$ is a conformal mapping from $\H$ 
976: onto some simply connected 
977: domain $\Omega$, 
978:  then we say that 
979: $( \Phi (K_t) , t \ge 0)$ is a chordal Schramm process
980: (with parameter $\kappa$) started from $\Phi(0)$
981: aiming at $\Phi(\infty)$ in $\Omega$. This process
982: is well-defined modulo increasing time-reparametrization
983: (i.e.,  if there exists a (random) continuous increasing $\psi$
984: such that for all $t$, $K_t= K_{\psi(t)}'$, then we say that the
985: two processes $K$ and $K'$ are equal).  
986: 
987: For the remainder of this paper, we will
988: assume that  $\kappa= 6 $ 
989: as this case  exhibits many very interesting properties.
990: It is possible to compute explicitely certain probabilities. For
991: instance, if $a,b>0$,  
992: $$
993: P ( T_{-a} < T_b ) 
994: = F (b/(a+b))
995: $$
996: where $F$ is the same as in Cardy's formula 
997: In fact, much more is true:
998: 
999: \begin {theorem}[\cite {LSW1}]
1000: \label {rest}
1001: Consider $SLE_6$ in the equilateral triangle, started from 
1002: $A$ and aiming at some point $M \in [B,C]$. Let $T$ denote the 
1003: first time $t$ at which $K_t \cap [B,C] \not= \emptyset$.
1004: Then, the law of $K_{T-} = \overline { \cup_{t<T} K_t}$ is 
1005: independent of the choice of $M \in [B,C]$, it satisfies 
1006: Cardy's formula and it is invariant under restriction. 
1007: \end {theorem}
1008: 
1009: \noindent
1010: Again, this property is only valid when $\kappa = 6$ and there exists a 
1011: more general version in terms of processes \cite {LSW1}.
1012: 
1013: 
1014: 
1015: \subsection {Link with reflected Brownian motion}
1016: 
1017: From Theorem \ref {rest}, Theorem  \ref {univ2} and the 
1018: strong Markov property,  it follows that 
1019: the chordal process $(K_t, t \ge 0)$ can be reinterpreted in terms
1020: of reflected Brownian motions. In particular, 
1021: the law of $K_{T-}$ is that of the $\T$-hull of 
1022: a stopped Brownian motion with oblique reflection. 
1023: More generally, finite-dimensional marginals of 
1024: the chordal $SLE_6$ process can
1025: for instance  be constructed as follows (other more general
1026: statements with other stopping times hold as well).
1027: 
1028: Suppose that $J_1, \ldots, J_p$ is a decreasing family  of 
1029: closed subsets of $\overline H$ such that $H \setminus 
1030: J_1, \ldots, H \setminus J_p$ are simply connected.
1031: When $(K_t, t \ge 0)$ is chordal $SLE_6$ in $\overline H$
1032: (started at $0$ and aiming at infinity), we
1033: put, for all $j \le p$,
1034: $$ T_j = \inf \{ t >0 \ : \ K_t \cap J_j \not= \emptyset \}.$$
1035: 
1036: Let $V$ denote a simply connected compact subset 
1037: of the upper half-plane such that $H'= \H \setminus V$
1038: is simply connected and let $x \in \partial H' $.
1039: Define a reflected Brownian motion $(B_s, s \ge 0)$ 
1040: in $H'$ with oblique reflection 
1041: angle (angle $\pi/3$ on the part of the boundary 
1042: between $x$ and $+\infty$ and reflection $2 \pi / 3$ 
1043: between $-\infty$ and $x$)
1044: as the conformal image in $H'$
1045: of reflected Brownian motion in $H$.
1046: If $J$ is a compact set, define
1047:  the stopping  time 
1048: $S=S (J)$ at which $
1049: V\cup B[0,S]$ intersects $J$ for the first time, and
1050: the hull $\V = \V ( V, x, J)$ of $V \cup B[0, S(J)]$ in $H$.
1051: 
1052: 
1053: Now, define recursively (using each time independent 
1054: Brownian motions), $V_0 = \{ 0\}$, $x_0 =  0$,
1055: $$
1056: V_{j+1} = \V ( V_j, x_j, J_j )
1057: \hbox { and }
1058: x_{j+1} = B ( S(J_j )).
1059: $$
1060: 
1061: \begin {theorem}
1062: \label {approx}
1063: The laws of $(V_{1}, \ldots, V_{p})$
1064: and of $(K_{T_1}, \ldots, K_{T_p})$ are identical.
1065: \end {theorem}
1066: 
1067: In particular, this implies that the law of $V_p$ is 
1068: identical to that of $K_{T_p}$ and therefore
1069: is independent of 
1070: $p$ and $J_1, \ldots, J_{p-1}$. This can be viewed 
1071: as  a way to
1072: reformulate the restriction property for $SLE_6$ in terms
1073: of reflected Brownian motions only.
1074: So far, there is no direct proof of this fact that does not
1075: use the link with Schramm's process.
1076: 
1077: \subsection {Relation between radial and chordal $SLE_6$}
1078: 
1079: Radial and chordal Schramm processes with parameter 6 
1080: are very closely related.
1081: The CCI property and the restriction property in their
1082: `process' versions can 
1083: be  very non-rigorously  described as follows:
1084: The evolution of $K_t$ (for radial and 
1085: chordal $SLE_6$) depends on $K_t$ only
1086: in a local way i.e.,  suppose that at time $t_0$,
1087: $K_t$ is increasing near some point $x \in \partial K_{t_0}$,
1088: then the evolution of $K_t$ immediately after $t_0$ depends only
1089: on how $K_{t_0}$ looks like in the neighbourhood of $x$
1090: (for a rigorous version of this statement, see \cite {LSW1,LSW2}).
1091: The link between radial and
1092: chordal $SLE_6$ (and this is only true when $\kappa=6$)
1093: is that 
1094:  this local evolution is the same for radial and
1095: chordal $SLE_6$, see \cite {LSW2}.
1096: This also leads to a description of the finite-dimensional 
1097: marginal
1098: laws of radial $SLE_6$ in terms
1099: of hulls of  reflected Brownian motion.
1100: 
1101: \section {Computation of the exponents}
1102: 
1103: We now give a very brief outline of the proof of 
1104: Theorem \ref {5/8}. 
1105:  As a consequence of the
1106: results stated  in the last 
1107: two sections, hulls of Brownian paths at certain 
1108: stopping
1109: times can be also constructed in an a priori completely
1110: different way using $SLE_6$. 
1111: The latter 
1112: turns out to be much better suited to compute 
1113: probabilities  involving only
1114: the shape of its complement, as $SLE_6$
1115: is  a process that is 
1116: `continuously growing to the outside', whereas  Brownian motion 
1117:  does incursions inside its own hull, so that 
1118: for instance the point at which the Brownian hull `grows' makes
1119: a lot of jumps. 
1120: 
1121: It is very easy to show that Theorem \ref{5/8}
1122: is a consequence of the following fact that we shall 
1123: now derive:
1124: Let $B$ and $B'$ denote two independent planar Brownian 
1125: motions started from $0$ and $\eps>0$, and
1126: killed when they hit the unit circle. Then, when $\eps \to 0$,
1127: \begin {equation}
1128: \label {5/4}
1129: P [ B \cap B' = \emptyset ] = \eps^{5/4 + o(1) }.
1130: \end {equation}
1131: To derive (\ref {5/4}), note first  that if $h_{K(B)}$
1132: denotes the harmonic measure of $\partial D$ 
1133: at $\eps$ in $D \setminus K(B)$ (here
1134: $K(B)$ is the hull of $B$), then 
1135: $$
1136: P [ B \cap B' = \emptyset | B ] 
1137: = h_{ K(B)}.
1138: $$
1139: Hence, using 
1140: Theorem \ref {univ}, 
1141: shows that 
1142: $$
1143: P [ B \cap B' = \emptyset ]
1144: = 
1145: E [ h_{K(B)} ]
1146: = 
1147: E [ h_{K_T} ]
1148: $$
1149: where $K_T$ is radial $SLE_6$ stopped 
1150: as in Theorem \ref {SLE6}.
1151: The construction
1152: of $K_T$ via Loewner's equation 
1153: makes it possible to study explicitly the 
1154: asymptotic behaviour of $E[ h_{K_T} ]$  
1155: when $\eps \to 0$ by computing 
1156: the highest eigenvalue computation of a differential 
1157: operator, see \cite {LSW2}. From this, 
1158: (\ref {5/4}) follows.
1159: 
1160: Similar (though more involved) arguments lead  
1161: to Theorem \ref {mandelbrot} and 
1162: to the determination of many other such critical exponents that 
1163: are defined in terms of planar Brownian paths
1164: \cite {LSW1,LSW2,LSW3,LSWa}.
1165: 
1166: \section {The conjecture for percolation}
1167: 
1168: Suppose that one performs critical bond percolation ($p=p_c=1/2$) 
1169: in the discrete half-plane $\Z \times \N$. Decide
1170: that all edges of the type $[x,x+1]$ when $x$ is on the real axis 
1171: are erased when $x \ge 0$ and open when $x<0$.
1172: Then, there exists a unique infinite cluster $C_-$ 
1173: formed by the negative half-axis and the union of all 
1174: clusters that are attached to it.
1175: We now explore the outer boundary of this cluster,
1176: starting from the point $(0,0)$. In other words, we 
1177: follow the 
1178: left-most possible  path that 
1179: is not allowed to cross open edges as shown in the picture 
1180: below (for convenience, we draw the line 
1181: that stays at `distance' $1/4$ from $C_-$).
1182:  The open edges are the thin plain lines, and the 
1183: exploration process is the thick plain line.
1184: 
1185: 
1186: \begin{figure}[h]
1187: \centering
1188: %\vspace {-2cm}
1189: \resizebox{0.7 \textwidth}{!} {\includegraphics {perco.eps}}
1190: %\vspace {-1cm}
1191: \caption{
1192: The discrete exploration process 
1193: }
1194: \end{figure}
1195: 
1196: 
1197: Note that this exploration process is almost symmetric
1198: because of the self-duality property of the planar lattice:
1199: It is (almost) the same than exploring the outer boundary of the  
1200: cluster of `open' edges in the dual lattice (i.e.
1201: duals of closed edges in the original 
1202: lattice - the thick dashed edges in the
1203: picture) attached to the `positive half-line'
1204: $(1-i)/ 2 + \N$
1205: (other choices of the lattice make this exploration process
1206: perfectly symmetric). 
1207: 
1208: 
1209: The conformal invariance 
1210: conjecture  leads naturally
1211: to the conjecture \cite {S,S2} that in the scaling
1212: limit, this  exploration process can be 
1213: described using Loewner's differential equation in 
1214: the upper half-plane, and that 
1215: the driving process $a(t)$ is a continuous symmetric
1216: Markov  process with stationary increments i.e. $a(t)
1217: = \sqrt {\kappa} W_t$ for some parameter $\kappa$. It is 
1218: then easy to identify $\kappa=6$ as the only possible candidate
1219: (looking for instance
1220: at the probability of crossing a square, see \cite {S2}), and the restriction
1221: property
1222: gives additional support to this conjecture.
1223: In particular, note that Cardy's formula for Schramm's
1224: process with parameter $6$ 
1225: would indeed correspond to the crossing probability.
1226: 
1227: So far, there is no mathematical proof of the 
1228: fact that this exploration process converges to Schramm's
1229: process.
1230: However, Theorem \ref {approx} and the fact that the exploration process 
1231: can be viewed as a discrete random walk reflected on its 
1232: past hull gives  
1233: at least a heuristic hand-waving justification.
1234:  
1235: One can also recover non-rigorously the exponents predicted
1236: for self-avoiding walks (see e.g. \cite {LW2}) using
1237: the exponents derived (rigorously) for Brownian motions
1238: and Schramm processes.
1239: 
1240: The only discrete models that have been mathematically shown 
1241: to exhibit some conformal invariance properties 
1242: in the scaling limit - apart from simple random walks - 
1243: are those studied by Kenyon in \cite {K1,K2,K3,K4}, namely
1244: loop-erased walks and uniform spanning trees. These are
1245: conjectured to correspond to 
1246: Schramm's processes with parameters 2 and 8, see \cite {S}.
1247: For partial results concerning percolation scaling
1248: limit
1249: and its conformal invariance, see \cite {AB,AB2,BS}.
1250: 
1251: For some other 
1252:  critical exponents related to Hausdorff dimensions
1253: of conformally invariant exceptional subsets of the planar 
1254: Brownian curve (such as pivoting cut points for
1255: instance) that we do not know (yet?) the value of, see   
1256: \cite {Bef}.
1257: 
1258: 
1259: \medbreak
1260: \noindent
1261: {\bf Acknowledgements.}
1262: Without my coauthors 
1263: Greg Lawler and Oded Schramm, 
1264: this paper would of course not exist.
1265: I also use this opportunity to 
1266: thank my Orsay  colleagues Richard Kenyon and
1267: Yves Le Jan for many very stimulating and inspiring discussions, as
1268: well as Vincent Beffara for the nice picture (Figure 1).
1269:  
1270:  
1271: \itemsep=\smallskipamount
1272: 
1273: \begin {thebibliography}{99}
1274: 
1275: \bibitem {Ai}{
1276: M. Aizenman,
1277: {\em The geometry of critical percolation and 
1278: conformal invariance}, Statphys19 (Xiamen, 1995), 104-120 (1996).}
1279: 
1280: \bibitem{AB} {
1281: M. Aizenman and A. Burchard, {\em  H\"older regularity
1282: and dimension bounds for random curves}, 
1283: Duke Math. J. {\bf 99}, 419-453 (1999).
1284: }
1285: 
1286: \bibitem{AB2} {
1287: M. Aizenman, A. Burchard, C. Newman, and D. Wilson,
1288: {\em Scaling limits for minimal and
1289:  random spanning trees in two dimensions},
1290: Random Str. Algo. {\bf 15}, 319-367 (1999).
1291: }
1292: 
1293: \bibitem {ADA} {
1294: M. Aizenman, B. Duplantier, A. Aharony,
1295: {\em Path crossing
1296:  exponents and the external perimeter in 2D percolation},
1297: Phys. Rev. Let. {\bf 83}, 1359-1362 (1999).}
1298: 
1299: \bibitem {Bef}
1300: {V. Beffara,
1301: {\em Conformally invariant subsets of the planar Brownian curve},
1302: preprint (2000).}
1303: 
1304: \bibitem {BPZ1}
1305: {A.A. Belavin, A.M. Polyakov, A.B. Zamolodchikov,
1306: {\em Infinite conformal symmetry of critical 
1307: fluctuations in two dimensions},
1308: J. Stat. Phys {\bf 34}, 763-774 (1984).}
1309: 
1310: \bibitem {BPZ2}
1311: {A.A. Belavin, A.M. Polyakov, A.B. Zamolodchikov,
1312: {\em Infinite conformal symmetry 
1313: in two-dimensional quantum field theory},
1314: Nucl. Phys. B {\bf 241}, 333-380 (1984).}
1315: 
1316: \bibitem {BS}
1317: {I. Benjamini, O. Schramm,
1318: {\em Conformal invariance of Voronoi percolation},
1319: Comm. Math. Phys. {\bf 197}, 75-107 (1998).}
1320: 
1321: \bibitem {Ca1}
1322: {J.L. Cardy, 
1323: {\em Conformal invariance and surface critical behavior},
1324: Nucl. Phys. B {\bf 240}, 514--532 (1984).}
1325: 
1326: \bibitem {Ca2}
1327: {J.L. Cardy,
1328: {\em Critical percolation in finite geometries},
1329: J. Phys. A {\bf 25}, L201-L206 (1992).}
1330: 
1331: \bibitem {Cabook}
1332: {J.L. Cardy,
1333: {\em Scaling and renormalization in Statistical 
1334: Physics}, Cambridge University Press, 1996.}
1335: 
1336:  \bibitem {Ca}
1337: {J.L. Cardy,
1338: {\em The number of incipient spanning clusters in two-dimensional
1339: percolation}, J. Phys. A {\bf 31}, L105 (1998).}
1340: 
1341: \bibitem {Dqg}
1342: {B. Duplantier,
1343: {\em Random walks and quantum gravity in two dimensions},
1344: Phys. Rev. Let. {\bf 82}, 5489-5492 (1998).}
1345: 
1346: \bibitem {Dcm}
1347: {B. Duplantier,
1348: {\em Two-dimensional copolymers and exact conformal multifractality},
1349: Phys. Rev. Let. {\bf 82}, 880--883 (1999).}
1350: 
1351: \bibitem {DK}{
1352: B. Duplantier, K.-H. Kwon,
1353: {\em Conformal invariance and intersection of random walks},
1354:  Phys. Rev. Let.
1355:  2514-2517 (1988).
1356: }
1357: 
1358: \bibitem {DS}
1359: {B. Duplantier, H. Saleur,
1360: {\em Exact determination of the percolation
1361: hull exponent in two dimensions},
1362: Phys. Rev. Lett. {\bf 58},
1363: 2325 (1987).}
1364: 
1365: \bibitem {Dur}
1366: {P. Duren,
1367: {\em Univalent functions}, Springer,  1983.}
1368: 
1369: \bibitem {Fl}
1370: {P.J. Flory,
1371: {\em The configuration of a real polymer chain},
1372: J. Chem. Phys {\bf 17}, 303-310 (1949).}
1373: 
1374: \bibitem{G}  
1375: {G. Grimmett, {\em Percolation},
1376: Springer-Verlag, 1989.
1377: }
1378: 
1379: \bibitem{ID} {
1380: C. Itzykon, J.-M. Drouffe,
1381:  {\em Statistical
1382: Field Theory}, Vol. 2, Cambridge University Press, 1989.
1383: }
1384:  \bibitem {K1}
1385: {
1386: R. Kenyon,
1387: {\em Conformal invariance of domino tiling}, Ann. Probab., to appear.}
1388:  
1389: \bibitem{K2}
1390: {
1391: R. Kenyon, 
1392: {\em The asymptotic determinant of the discrete
1393: Laplacian}, preprint (1998).
1394: }
1395: 
1396: \bibitem {K3}
1397: {R. Kenyon,
1398: {\em Long-range properties of spanning trees in $\Z^2$},
1399: J. Math. Phys., to appear.}
1400:  
1401: \bibitem {K4}
1402: {R. Kenyon,
1403: {\em Dominos and the Gaussian free field}, 
1404: preprint (2000).
1405: }
1406:  
1407:  \bibitem {LPS}
1408: {R. Langlands, Y. Pouillot, Y. Saint-Aubin,
1409: {\em Conformal invariance in two-dimensional percolation},
1410: Bull. A.M.S. {\bf 30}, 1--61, (1994).}
1411: 
1412: \bibitem {Lhaus}
1413: {G.F. Lawler,
1414: {\em The dimension of the frontier of planar Brownian motion},
1415: Electron. Comm. Probab. {\bf 1} 29-47 (1996).}
1416: 
1417: \bibitem{Lbuda}
1418: {
1419: G.F.  Lawler, 
1420: {\em Geometric and fractal properties of Brownian motion
1421: and random walks paths in two and three dimensions}, in
1422: {\em Random Walks, Budapest 1998}, Bolyai
1423: Society Mathematical Studies {\bf 9}, 219--258 (1999)
1424: .}
1425: 
1426: \bibitem {LP}
1427: { G.F. Lawler, E.E. Puckette,
1428: {\em The intersection exponent for simple random walk},
1429: Comb. Probab. Comput., to appear.}
1430: 
1431: \bibitem {LSW1}
1432: {G.F. Lawler, O. Schramm, W. Werner,
1433: {\em  Values of Brownian   
1434:              intersection exponents I:  
1435:            Half-plane exponents}, preprint (1999). 
1436: }
1437: 
1438: \bibitem {LSW2}
1439: {G.F. Lawler, O. Schramm, W. Werner,
1440: {\em Values of Brownian intersection exponents II: Plane exponents},
1441: preprint (2000).
1442: }
1443: 
1444: \bibitem {LSW3}
1445: {G.F. Lawler, O. Schramm, W. Werner,
1446: {\em Values of Brownian intersection exponents III:
1447: Two-sided exponents},
1448: preprint (2000).
1449: }
1450: 
1451: \bibitem {LSWa}
1452: {G.F. Lawler, O. Schramm, W. Werner,
1453: {\em Analyticity of planar Brownian intersection exponents},
1454:               preprint (2000).
1455: }
1456: 
1457: %\bibitem {LSWip}
1458: %{G. F. Lawler, O. Schramm, W. Werner (2000),
1459: %{\em Critical percolation and reflected Brownian motion},
1460: % in preparation.}
1461: 
1462: \bibitem {LW1}
1463: {G.F. Lawler, W. Werner,
1464: {\em Intersection exponents for planar Brownian motion},
1465: Ann. Probab. {\bf 27}, 1601-1642 (1999).}
1466: 
1467:  \bibitem {LW2}
1468: {G.F. Lawler, W. Werner,
1469: {\em Universality for conformally invariant intersection exponents},
1470: J. European Math. Soc., to appear.}
1471:  
1472: 
1473: 
1474: \bibitem {MS} {N. Madras, G. Slade,
1475: {\em The self-avoiding walk}, Birkh\"auser, Boston, 1993.}
1476: 
1477: 
1478: \bibitem {M} {B.B. Mandelbrot, 
1479: {\em The fractal geometry of nature,}
1480: Freeman, 1982.}
1481: 
1482: 
1483: \bibitem {N1}
1484: {B. Nienhuis,
1485: {\em Critical behavior of two-dimensional spin
1486: models and charge asymmetry in the Coulomb gas},
1487: J. Stat. Phys. {\bf 34}, 731--761 (1983).}
1488: 
1489: \bibitem {S}{
1490: O. Schramm, 
1491: {\em Scaling limits of loop-erased random walks and
1492: uniform spanning trees}, Israel J. Math, to appear.}
1493: 
1494: \bibitem {S2}{
1495: O. Schramm, 
1496: {\em Conformal invariant scaling limits},
1497: in preparation.}
1498: 
1499: \bibitem {VW}{
1500: S.R.S. Varadhan, R.J. Williams,
1501: {\em Brownian motion in a wedge with oblique reflection},
1502: Comm. Pure Appl. Math. {\bf 38} 405-443 (1985).}
1503: 
1504: \end {thebibliography}
1505: 
1506: 
1507: 
1508: Laboratoire de Math\'ematiques,
1509: 
1510: Universit\'e Paris-Sud,
1511: 
1512: B\^at. 425
1513: 
1514: 91405 Orsay cedex, France
1515: 
1516: wendelin.werner@math.u-psud.fr
1517: 
1518: 
1519: \end{document}
1520: