Conformal Loop Ensembles (CLE)

Updated 22 April 2026

Conformal Loop Ensembles (CLE) are probability measures on collections of non-crossing loops in planar domains, defined by conformal invariance and the domain Markov property.
CLE distinguishes between simple (κ ≤ 4) and non-simple (κ > 4) regimes, each exhibiting distinct geometric and probabilistic characteristics.
CLE unifies the scaling limits of critical 2D models—such as O(n), Potts, and random cluster models—and connects with SLE, Brownian loop-soups, and Liouville quantum gravity.

A conformal loop ensemble (CLE) is a canonical probability measure on random collections of non-crossing loops in a planar domain that satisfy conformal invariance and a domain Markov property. Parameterized by $\kappa \in (8/3,8)$ , CLE unifies the scaling limits of interfaces in numerous two-dimensional critical models, including the O( $n$ ) loop, Potts, and random cluster models. CLEs are realized for all $\kappa \in (8/3,8)$ , with sharp distinctions between the "simple" regime ( $\kappa \leq 4$ ) and the "non-simple" regime ( $\kappa > 4$ ), each exhibiting distinct probabilistic and geometric characteristics. CLEs are fundamentally connected to Schramm–Loewner Evolution (SLE), Brownian loop-soups, Liouville quantum gravity, integrable models in conformal field theory, and random planar maps.

1. Definitions, Constructions, and Fundamental Properties

A CLE $_\kappa$ in a simply connected domain $D\subset\mathbb{C}$ is a random, countable collection $\Gamma$ of non-crossing loops characterized by:

Conformal invariance: If $\varphi:D\to D'$ is conformal, then $\varphi(\Gamma)$ has the law of a CLE $n$ 0 in $n$ 1.
Domain Markov (restriction) property: After conditioning on the outermost loops that intersect a set $n$ 2, the interiors of these loops (and the exterior minus the union of these loops) are filled with independent CLE $n$ 3 in each component (Sheffield et al., 2010).
Finiteness: For any compact $n$ 4, only finitely many loops of diameter larger than any given $n$ 5 intersect $n$ 6 (Miller et al., 2022).

Two Regimes

Simple CLE ( $n$ 7): Loops are disjoint, simple, and do not touch the boundary.
Non-simple CLE ( $n$ 8): Loops may touch themselves and each other, and can intersect the boundary, but remain non-crossing (Sheffield et al., 2010, Gwynne et al., 2018).

Main Constructions

Branching SLE $n$ 9: Explore SLE $\kappa \in (8/3,8)$ 0-type branches targeted at dense sets; loops are formed as SLE excursions (Sheffield et al., 2010, Gwynne et al., 2018).
Brownian loop-soup: For $\kappa \in (8/3,8)$ 1, the outer boundaries of clusters in a Poissonian Brownian loop-soup of intensity $\kappa \in (8/3,8)$ 2 are exactly the loops of CLE $\kappa \in (8/3,8)$ 3 (Sheffield et al., 2010, Kemppainen et al., 2014).

Domain Extensions and Möbius Invariance

CLEs can be defined in doubly connected domains (annuli), punctured discs, and the whole-plane (Riemann sphere), with exact constructions and Markov properties carrying over (including inversion invariance for both simple and non-simple regimes) (Gwynne et al., 2018, Kemppainen et al., 2014, Sheffield et al., 2014). The nested version is constructed by iterating the basic CLE in each connected component.

Table: CLE Constructions by Regime

$\kappa \in (8/3,8)$ 4	Loop Type	Construction
$\kappa \in (8/3,8)$ 5	simple	Branching SLE, Loop-soup, Restriction axioms
$\kappa \in (8/3,8)$ 6	non-simple	Branching SLE, Limiting version of loop-soup

2. Fractal Geometry: Carpets, Gaskets, and Dimensions

Given a CLE $\kappa \in (8/3,8)$ 7 sample $\kappa \in (8/3,8)$ 8 in $\kappa \in (8/3,8)$ 9, define the carpet ( $\kappa \leq 4$ 0) or gasket ( $\kappa \leq 4$ 1):

$\kappa \leq 4$ 2

Key geometric properties:

Hausdorff dimension: Almost surely,

$\kappa \leq 4$ 3

which agrees with Duplantier–Saleur’s predictions for the scaling limits of the O( $\kappa \leq 4$ 4) and FK random cluster models (Miller et al., 2012, Miller et al., 2022).

For $\kappa \leq 4$ 5, the carpet is a fractal measure-zero Cantor set; for $\kappa \leq 4$ 6, the gasket includes dense self-touching loops and is topologically subtle.
Extreme nesting: For the set of points $\kappa \leq 4$ 7 where the number of loops surrounding $\kappa \leq 4$ 8 grows at asymptotic rate $\kappa \leq 4$ 9, the almost-sure Hausdorff dimension is $\kappa > 4$ 0, with explicit rate function $\kappa > 4$ 1 determined via large deviations (Miller et al., 2013).

3. Metric, Volume Measure, and Diffusions on the CLE Carpet/Gasket

A strong program aims to rigorously define the natural metric and measure structure on CLE carpets and gaskets:

Canonical Metric

For $\kappa > 4$ 2 ("simple" regime), approximations to the chemical distance (infimum of $\kappa > 4$ 3-neighborhood area over paths avoiding all loops) are tight and Hölder continuous; there is a unique subsequential scaling limit in the Arzelà–Ascoli topology, conjecturally conformally covariant (Miller, 2021).
For $\kappa > 4$ 4 ("non-simple" regime), a geodesic metric can be defined using the infimum of the Euclidean diameter over admissible paths avoiding all loops. Uniqueness and conformal covariance of this metric are established, along with locality, Markov, and scaling properties (Miller et al., 14 Aug 2025). The metric is almost surely a deterministic function of the CLE realization.

Conformally Covariant Volume Measure

There exists a unique, up-to-scalar, conformally covariant and Markovian measure $\kappa > 4$ 5 supported on the CLE carpet/gasket:

$\kappa > 4$ 6

This measure arises as a deterministic function of the CLE and can be constructed explicitly via Liouville quantum gravity and multiplicative cascades (Miller et al., 2022, Miller et al., 2020).

Canonical Diffusion (Brownian Motion)

For $\kappa > 4$ 7, there is a unique diffusion process ("CLE $\kappa > 4$ 8-Brownian motion") on the gasket whose Dirichlet form is a locally determined resistance form, translation-invariant, scale-covariant, and conformally covariant (Miller et al., 4 Dec 2025). The scaling limit of discrete random walks on models converging to CLE $\kappa > 4$ 9 is conjectured to yield this continuum process.

Table: CLE Carpet/Gasket Metric and Measure Structures

Structure	Regime	Existence	Properties
(Geodesic) metric	Simple/Non-simple	Proven	Conformal, unique
Canonical measure	All $_\kappa$ 0	Proven	Conformal exponent $_\kappa$ 1
Brownian motion	Non-simple	Proven	Uniqueness, resistance

4. CLE in Statistical Mechanics: Critical Models and Connection Probabilities

CLE precisely describes scaling limits of interfaces in critical 2D models:

O( $_\kappa$ 2) loop models: For $_\kappa$ 3, the dilute phase corresponds to simple CLE $_\kappa$ 4; critical thresholds match crossing and connection probabilities (Miller et al., 2017, Miller et al., 2013).
FK random-cluster models: For $_\kappa$ 5, the scaling limit of critical FK cluster boundaries is CLE $_\kappa$ 6 ( $_\kappa$ 7) (Cai, 30 Mar 2026, Liu et al., 2024).
Percolation and Ising: For $_\kappa$ 8 (percolation) and $_\kappa$ 9 (Ising-FK), full four-point and one-arm exponent formulas are rigorously established (Cai, 30 Mar 2026, Liu et al., 2024).

Connection probabilities

For conformal rectangles with alternate wired/free boundaries, the probability that the two wired arcs are joined via a single loop is

$D\subset\mathbb{C}$ 0

A universal formula, matching those in discrete models and providing a dictionary between $D\subset\mathbb{C}$ 1 and the physical parameters $D\subset\mathbb{C}$ 2 or $D\subset\mathbb{C}$ 3 (Miller et al., 2017).

Arm Exponents

CLE provides rigorous derivations and transcendentally explicit values for (bulk and boundary) arm exponents in the scaling limit of percolation/FK/Potts models. For the bulk one-arm exponent in CLE $D\subset\mathbb{C}$ 4 percolations, the exact value is determined via conformal block computations and Liouville CFT structure constants, e.g., $D\subset\mathbb{C}$ 5 in the 3-state Potts model (Liu et al., 2024).

5. CLE, SLE, and Integrability: Couplings, Dualities, and Observables

CLE and SLE are deeply intertwined:

CLE as branching, target-invariant SLE $D\subset\mathbb{C}$ 6 trees: For $D\subset\mathbb{C}$ 7, both the simple and non-simple regimes (Sheffield et al., 2010).
Coupling with the Gaussian Free Field: CLE $D\subset\mathbb{C}$ 8 emerges as level lines of the GFF, with i.i.d. sign labels; SLE–GFF couplings extend to all $D\subset\mathbb{C}$ 9, providing a route to construct percolation interfaces in CLE carpets (Doyon, 2012, Miller et al., 2016).
Continuum Edwards–Sokal coupling: There is a precise duality between simple and non-simple CLEs via random coloring and cluster boundary tracing, matching the coupling between FK and Potts models (Miller et al., 2016, Miller et al., 2020).

Integrable Structure and Exact Observables

Three-point and multi-point connectivity: Certain nontrivial three-point observables (e.g., three-point cluster or nesting Green's functions) in CLE are predicted and, for percolation, proved to be given exactly by variants of the imaginary DOZZ formula, connecting CLE to Liouville theory and CFT (Ang et al., 2021).
Boundary connectivity (BPZ/PDE methods): For $\Gamma$ 0, boundary four-point connectivities of CLE are solutions to explicit third-order ODEs derived from BPZ equations, furnishing exact formulas and revealing logarithmic singularities associated to logarithmic CFTs (FK-Ising) (Cai, 30 Mar 2026).

6. CLE on Liouville Quantum Gravity and Stable Trees

Decorating an independent Liouville quantum gravity (LQG) surface by CLE introduces deep connections to stable growth–fragmentation processes and Poissonian ensembles of quantum surfaces (“quantum disks”). Under Markovian exploration of CLE on LQG,

The related quantum surface decomposition forms a growth–fragmentation tree whose genealogical (mass) evolution matches a $\Gamma$ 1-stable Lévy process (Miller et al., 2020).
These descriptions rigorously match the scaling limits of decorated random planar maps with large faces, unifying LQG, CLE, and discrete models through stable process and welding constructions.

Table: CLE and LQG Correspondence

CLE parameter	LQG $\Gamma$ 2	Stable process	Loop density
$\Gamma$ 3	$\Gamma$ 4	index $\Gamma$ 5	$\Gamma$ 6

7. Nesting Field, Extreme Value Statistics, and Gaussian Free Field

CLE supports a "nesting field": the spatial field of fluctuations in the number of loops surrounding a small ball, which in distributional sense converges (for $\Gamma$ 7, with $\Gamma$ 8 Bernoulli weights) to the Gaussian free field (GFF). In other cases ( $\Gamma$ 9), the nesting field is non-Gaussian but retains conformal invariance, with precise laws derived for moments, covariance, and joint statistics (Miller et al., 2013). The tools connect CLE nesting to extreme-value statistics and thick point theory of the continuum GFF.

References

(Sheffield et al., 2010) "Conformal Loop Ensembles: Construction via Loop-soups"
(Miller et al., 2013) "Extreme nesting in the conformal loop ensemble"
(Miller et al., 2013) "The conformal loop ensemble nesting field"
(Miller et al., 2012) "The Hausdorff dimension of the CLE gasket"
(Miller, 2021) "Tightness of approximations to the chemical distance metric for simple conformal loop ensembles"
(Miller et al., 14 Aug 2025) "Existence and uniqueness of the conformally covariant geodesic metric on non-simple conformal loop ensemble gaskets"
(Miller et al., 4 Dec 2025) "Existence and uniqueness of the canonical Brownian motion in non-simple conformal loop ensemble gaskets"
(Miller et al., 2022) "Existence and uniqueness of the conformally covariant volume measure on conformal loop ensembles"
(Miller et al., 2020) "Simple Conformal Loop Ensembles on Liouville Quantum Gravity"
(Cai, 30 Mar 2026) "Boundary four-point connectivities of conformal loop ensembles"
(Liu et al., 2024) "The bulk one-arm exponent for the CLE $\varphi:D\to D'$ 0 percolations"
(Ang et al., 2021) "Integrability of Conformal Loop Ensemble: Imaginary DOZZ Formula and Beyond"
(Miller et al., 2016) "CLE percolations"
(Gwynne et al., 2018, Kemppainen et al., 2014, Sheffield et al., 2014) (full-plane/nested CLEs and Möbius invariance)
(Miller et al., 2017) "Connection probabilities for conformal loop ensembles"

These works collectively establish CLE as the universal scaling limit in two-dimensional critical phenomena, with full identification of its metric, measure, fractal, and integrable structures.