[Paper Review] Exploration trees and conformal loop ensembles

Scott Sheffield⋆|ArXiv.org|Sep 6, 2006

Stochastic processes and statistical mechanics8 references8 citations

TL;DR

This paper introduces and constructs the conformal loop ensemble CLE(κ) for 8/3 ≤ κ ≤ 8 using branching SLE processes known as exploration trees, establishing a continuum limit for random loop models in statistical mechanics. The key contribution is a rigorous construction of CLE(κ) via conformal invariance and Markov properties, with CLE(6) identified as the scaling limit of site percolation on the triangular lattice.

ABSTRACT

We construct and study the conformal loop ensembles CLE(kappa), defined for all kappa between 8/3 and 8, using branching variants of SLE(kappa) called exploration trees. The conformal loop ensembles are random collections of countably many loops in a planar domain that are characterized by certain conformal invariance and Markov properties. We conjecture that they are the scaling limits of various random loop models from statistical physics, including the O(n) loop models.

Motivation & Objective

• To define and construct the conformal loop ensemble CLE(κ) for 8/3 ≤ κ ≤ 8 using continuum exploration trees derived from SLE(κ;ρ) processes.
• To establish that CLE(κ) arises as the scaling limit of discrete loop models such as the O(n) loop model and site percolation on the triangular lattice.
• To prove uniqueness and symmetry properties of CLE(κ) under conformal invariance and time reversal for 4 < κ < 8.
• To formulate conjectures on the scaling limits of Potts models, FK clusters, and height functions, and their relation to the Gaussian free field.
• To explore the duality between CLE(κ) and CLE(16/κ), particularly in the context of outermost spin clusters in the q-state Potts model.

Proposed method

• Constructs CLE(κ) by defining a continuum exploration tree via radial and chordal SLE(κ;ρ) processes, with ρ = κ − 6.
• Uses Bessel processes and Lévy skew stable processes to model the radial SLE(κ;κ−6) process, linking them to the branching structure of exploration trees.
• Applies conformal invariance and the Markov property to characterize CLE(κ) as the unique random collection of loops satisfying these symmetries.
• Establishes a one-to-one correspondence between exploration trees and loop ensembles in the continuum, enabling the construction of CLE(κ) from the tree's branching structure.
• Introduces the concept of the CLE gasket and derives its conformal radius distribution, crucial for characterizing the loop structure.
• Proposes a duality conjecture: the outermost spin cluster in the q-state Potts model for wired boundary conditions corresponds to CLE(16/κ) when the original CLE is CLE(κ).

Experimental results

Research questions

• RQ1 Is the scaling limit of the loop boundaries in the critical O(n) loop model on the hexagonal lattice given by CLE(κ) for some κ?
• RQ2 Does the time reversal symmetry of chordal SLE(κ;κ−6) hold for 4 < κ < 8, and if so, does it imply symmetry in CLE(κ)?
• RQ3 Is the scaling limit of the FK cluster model for the q-state Potts model at criticality given by CLE(κ), where q = 2 + 2cos(8π/κ) and 4 ≤ κ ≤ 8?
• RQ4 Can a continuum analog of the FK cluster expansion be defined for CLE(κ), and does it yield a consistent scaling limit for Potts models?
• RQ5 Is the law of the outermost spin cluster in the q-state Potts model (wired boundary conditions) equivalent to CLE(16/κ) for κ ∈ [4,6) and q ∈ (1,4]?

Key findings

• CLE(8) is almost surely a single space-filling loop, corresponding to the scaling limit of the outer boundary of the uniform spanning tree.
• CLE(8/3) almost surely contains no loops, indicating a phase transition at the lower end of the κ range.
• For 8/3 < κ < 8, CLE(κ) consists of almost surely countably infinite simple loops, with loop intersections with the boundary occurring iff κ > 4.
• When κ = 6, CLE(6) is equivalent to the scaling limit of cluster boundaries in site percolation on the triangular lattice.
• For 4 < κ < 8, any random loop ensemble satisfying conformal invariance and the Markov property (with positive probability of boundary intersection) must be a CLE(κ).
• The paper conjectures that the outermost spin cluster in the q-state Potts model for wired boundary conditions corresponds to CLE(16/κ), suggesting a duality between CLE(κ) and CLE(16/κ).