RésuméIn this research announcement, we show that SLE curves can in fact be viewed as boundaries of certain clusters of Brownian loops (of the clusters in a Brownian loop soup). For small densities c of loops, we show that the outer boundaries of the clusters created by the Brownian loop soup are SLEκ-type curves where κ ∈ (8/3, 4] and c related by the usual relation c = (3κ − 8)(6 − κ)/2κ (i.e. c corresponds to the central charge of the model). This gives (for any Riemann surface) a simple construction of a natural countable family of random disjoint SLEκ loops, that behaves "nicely" under perturbation of the surface and is related to various aspects of conformal field theory and representation theory.
BackgroundThe goal of this paper is announce some results that relate the Brownian loop soup introduced in [16] to the Schramm-Loewner Evolutions (SLE) (in particular for values of the parameter between 8/3 and 4) and random families of disjoint SLE loops (as they might appear in the scaling limit of many 2d statistical physics models, and in conformal field theory). A more complete paper [26] on the same subject (with proofs, that discusses also various consequences of this approach) is in preparation.SLE processes have been introduced by Schramm in [21], building on the observation that they are the only processes that have a certain (conformal) Markovian-type property. There is a one-dimensional family of SLEs (indexed by a positive real parameter κ), and they are the only possible candidates for the scaling limits of interfaces for two-dimensional critical systems that are believed to be conformally invariant. This definition of SLE via Loewner's equation is a dynamic one-dimensional construction : One basically describes the law of η[t, t+ dt] given η [0, t], and integrates this with respect to t. See e.g. [21,20,10,23] for an introduction to (chordal) SLE. Furthermore, it is worthwhile stressing that this construction describes one interface, corresponding to specific boundary conditions in the discrete model, but that it does in general not give immediately access to the "complete scaling limit" of the system. This raises the following question : Is there a simple and natural way to define at once a whole family 1 of SLE loops in a domain that might describe simultaneously all boundaries of clusters ?In [14], a different characterization of the SLE 8/3 random curve was derived. It is shown to be the unique random curve in a domain, that satisfies a certain conformal restriction property. This characterization is "global" and does not use (directly) the Markovian property. It also enabled to identify this curve with the outer boundary of a certain reflected Brownian motion [14], and with the outer boundary of a certain union of Brownian excursions [25]. Hence, it is geometrically possible to construct SLE 8/3 from planar Brownian motions (recall also that SLE 8/3 is conjectured to be the scaling limit of the half-plane self-avoiding walk [13]).When κ = 2, SLE has been proved in [12] to be the scaling limit ...