Schramm–Loewner evolution

In probability theory, the Schramm–Loewner evolution is a conformally invariant stochastic process. It is a family of random planar curves that are generated by solving Charles Loewner's differential equation with Brownian motion as input. It was discovered by Oded Schramm (2000) as a conjectured scaling limit of the planar uniform spanning tree (UST) and the planar loop-erased random walk (LERW) probabilistic processes, and developed by him together with Greg Lawler and Wendelin Werner in a series of joint papers. There are variations on the name: evolution is sometimes replaced with equation, and Loewner is often spelled Löwner. In the initial papers on the topic "Schramm" was replaced with "stochastic", but since the passing away of Oded Schramm, this variation is not used a lot. The acronym SLE is perhaps the most popular way to address this object.

Besides UST and LERW, the Schramm–Loewner evolution is conjectured or proved to describe the scaling limit of various stochastic processes in the plane, such as critical percolation, the critical Ising model, the dimer model, and other critical statistical mechanics models that exhibit conformal invariance. The SLE curves are the scaling limits of interfaces and other non-self-intersecting random curves in these models. The main idea is that the conformal invariance and a certain Markov property inherent in such stochastic processes together make it possible to encode these planar curves into a one-dimensional Brownian motion running on the boundary of the domain (the driving function in Loewner's differential equation). This way, many important questions about the planar models can be translated into exercises in Ito calculus. Indeed, several mathematically non-rigorous predictions made by physicists using conformal field theory have been proved using this strategy.

Contents

The Loewner equation

Main article: Loewner differential equation

If D is a simply connected, open complex domain not equal to C, and γ is a simple curve in D starting on the boundary (a continuous function with γ(0) on the boundary and γ((0, ∞)) in D), then for each t ≥ 0, the complement Dt of γ([0, t]) is simply connected and therefore conformally isomorphic to D by the Riemann mapping theorem. If ƒt is a suitable normalized isomorphism from D to Dt, then it satisfies a differential equation found by Loewner (1923, p. 121) in his work on the Bieberbach conjecture. Sometimes it is more convenient to use the inverse function gt of ƒt, which is a conformal mapping from Dt to D.

In Loewner's equation, z is in the domain D, t ≥ 0, and the boundary values at time t=0 are ƒ0(z) = z or g0(z) = z. The equation depends on a driving function ζ(t) taking values in the boundary of D. If D is the unit disk and the curve γ is parameterized by "capacity", then Loewner's equation is

\frac{\partial f_t(z)}{\partial t} = -z f^\prime_t(z)\frac{\zeta(t)%2Bz}{\zeta(t)-z}   or   \frac{\partial g_t(z)}{\partial t} = g_t(z)\frac{\zeta(t)%2Bg_t(z)}{\zeta(t)-g_t(z)}.

When D is the upper half plane the Loewner equation differs from this by changes of variable and is

\frac{\partial f_t(z)}{\partial t} = \frac{ 2f_t^\prime(z)}{\zeta(t)-z}   or   \frac{\partial g_t(z)}{\partial t} = \frac{ 2}{g_t(z)-\zeta(t)}.

The driving function ζ and the curve γ are related by

\displaystyle f_t(\zeta(t)) = \gamma(t)   or   \displaystyle \zeta(t) = g_t(\gamma(t))

where ƒt and gt are extended by continuity.

Example

If D is the upper half plane and the driving function ζ is identically zero, then

f_t(z) = \sqrt{z^2-4t}
g_t(z) = \sqrt{z^2%2B4t}
\gamma(t) = 2i\sqrt{t}
D_t is the upper half plane with the line from 0 to 2i\sqrt{t} removed.

Schramm-Loewner evolution

Schramm–Loewner evolution is the random curve γ given by the Loewner equation as in the previous section, for the driving function

\displaystyle\zeta(t)=B(\kappa t)

where B(t) is Brownian motion on the boundary of D, scaled by some real κ. In other words Schramm–Loewner evolution is a probability measure on planar curves, given as the image of Wiener measure under this map.

In general the curve γ need not be simple, and the domain Dt is not the complement of γ([0,t]) in D, but is instead the unbounded component of the complement.

There are two versions of SLE, using two families of curves, each depending on a non-negative real parameter κ:

SLE depends on a choice of Brownian motion on the boundary of the domain, and there are several variations depending on what sort of Brownian motion is used: for example it might start at a fixed point, or start at a uniformly distributed point on the unit circle, or might have a built in drift, and so on. The parameter κ controls the rate of diffusion of the Brownian motion, and the behavior of SLE depends critically on its value.

The two domains most commonly used in Schramm–Loewner evolution are the upper half plane and the unit circle. Although the Loewner differential equation in these two cases look different, they are equivalent up to changes of variables as the unit circle and the upper half plane are conformally equivalent. However a conformal equivalence between them does not preserve the Brownian motion on their boundaries used to drive Schramm–Loewner evolution.

Special values of κ

When SLE corresponds to some conformal field theory, the parameter κ is related to the central charge c of the conformal field theory by

c = \frac{(8-3\kappa)(\kappa-6)}{2\kappa}.

Each value of c < 1 corresponds to two values of κ, one value κ between 0 and 4, and a "dual" value 16/κ greater than 4.

Beffara (2008) showed that the Hausdorff dimension of the paths (with probability 1) is equal to min(2, 1 + κ/8).

Applications

Lawler, Schramm & Werner (2001) used SLE6 to prove the conjecture of Mandelbrot (1982) that the boundary of planar Brownian motion has fractal dimension 4/3.

Critical percolation on the triangular lattice was proved to be related to SLE with \kappa =6 by Stanislav Smirnov.[1] Combined with earlier work of Harry Kesten,[2] this led to the determination of many of the critical exponents for percolation.[3] This breakthrough, in turn, allowed to further analyze many aspects of this model.[4][5]

Loop-erased random walk was shown to converge to SLE with \kappa=2 by Lawler, Schramm and Werner.[6] This allowed to derive many quantitative properties of loop-erased random walk (some of which were derived earlier by Richard Kenyon[7]). The related random Peano curve outlining the uniform spanning tree was shown to converge to SLE with \kappa=8.[6]

References

  1. ^ Stanislav smirnov, Critical percolation in the plane, preprint (2001), http://arxiv.org/abs/0909.4499
  2. ^ Harry Kesten, Scaling relations for 2D-percolation. Comm. Math. Phys. 109 (1987), no. 1, 109--156.
  3. ^ Stanislav Smirnov and Wendelin Werner, Critical exponents for two-dimensional percolation. Math. Res. Lett. 8 (2001), no. 5-6, 729--744
  4. ^ Oded Schramm and Jeffrey E. Steif, Quantitative noise sensitivity and exceptional times for percolation. Ann. of Math. (2) 171 (2010), no. 2, 619--672
  5. ^ Christophe Garban, Gábor Pete and Oded Schramm, Pivotal, cluster and interface measures for critical planar percolation, preprint (2010), http://arxiv.org/abs/1008.1378
  6. ^ a b Gregory F. Lawler, Oded Schramm and Wendelin Werner, Conformal invariance of planar loop-erased random walks and uniform spanning trees. Ann. Probab. 32 (2004), no. 1B, 939--995.
  7. ^ Richard Kenyon, Long range properties of spanning trees. J. Math. Phys. 41:3, 1338--1363 (2000)

External links