OctoInformBeta

OpenAI preprint, September 2026 · §1. Introduction · §1

Theorem 1.1 [Brownian continuum random tree limits of finite Fortuin-Kasteleyn maps above four]: For every q>4q > 4, rescaled metric measure spaces of maps converge in distribution to the Brownian tree

Fortuin-Kasteleyn planar maps converge to the Brownian continuum random tree.

Hover over a blue or green term in the text, or a node in the graph, to find it in the other. Select it to read its declaration both in Lean and in words.

The Setting

When decorated by Fortuin-Kasteleyn percolation with cluster weight q>4q > 4, large maps undergo geometric degeneration: their typical macroscopic geometry flattens into tree-like structures. For a planar map of size nn, rescaling the graph distance by a deterministic scaling constant c(q)/n+1c(q) / \sqrt{n+1} and equipping the map with its normalized degree measure produces a random compact metric measure space. As the map size nn tends to infinity, this random metric measure space converges in distribution to the Brownian continuum random tree.

A planar map of size nn is represented as a rotation system on 2(n+1)2(n+1) darts, indexed by the set Dn={0,…,2n+1}D_n = \{0, \dots, 2n+1\}. A labelled planar map (a,s)(a, s) on DnD_n is a pair of permutations of DnD_n satisfying two geometric conditions: the spanning graph is connected, and Euler’s formula |Orbit(s)|+|Orbit(sa)|=n+3|\mathrm{Orbit}(s)| + |\mathrm{Orbit}(s a)| = n + 3 holds for the fixed-point-free involution aa and vertex rotation ss.✦ An edge of a labelled planar map (a,s)(a, s) on DnD_n is an orbit of the dart reversal involution aa, and a vertex of (a,s)(a, s) is an orbit of the vertex rotation ss. A rooted planar map MM is an equivalence class of labelled maps modulo root-preserving dart relabelling, where an isomorphism is a permutation of DnD_n fixing dart 00 and conjugating both aa and ss.✦

Let ℳn\mathcal{M}_n denote the set of rooted planar maps of size nn. Given a map M∈ℳnM \in \mathcal{M}_n with vertex set V(M)V(M), edge set E(M)E(M), and graph distance dMd_M, and an edge subset A⊆E(M)A \subseteq E(M), the spanning subgraph (V(M),A)(V(M), A) has k(M,A)k(M, A) connected components. The Fortuin-Kasteleyn weight wq(M,A)w_q(M, A) of the pair (M,A)(M, A) at Potts parameter q>4q > 4 is wq(M,A)=qk(M,A)+|A|−|V(M)|2,w_q(M, A) = q^{k(M, A) + \frac{|A| - |V(M)|}{2}},(1)✦ and the finite partition function Zn(q)Z_n(q) is the sum over all configurations, Zn(q)=∑M∈ℳn∑A⊆E(M)wq(M,A).Z_n(q) = \sum_{M \in \mathcal{M}_n} \sum_{A \subseteq E(M)} w_q(M, A).(2)✦ The normalized degree measure μM\mu_M on V(M)V(M) is the probability measure on V(M)V(M) that assigns mass degM(v)/(2(n+1))\mathrm{deg}_M(v) / (2(n+1)) to each vertex v∈V(M)v \in V(M), defined by pushing forward the uniform measure on darts under the projection to vertex orbits.✦

A metric measure presentation X=(X,dX,ℬX,μX)X = (X, d_X, \mathcal{B}_X, \mu_X) consists of a set XX equipped with a distance function dXd_X, a Borel σ\sigma-algebra ℬX\mathcal{B}_X, and a finite Borel measure μX\mu_X.✦ A valid metric measure presentation X=(X,dX,ℬX,μX)X = (X, d_X, \mathcal{B}_X, \mu_X) is a metric measure presentation such that (X,dX)(X, d_X) is a compact metric space and μX\mu_X is a probability measure.✦ Given two valid metric measure presentations XX and YY, an isometric amalgamation of XX and YY is a pseudometric space on the disjoint union X⊕YX \oplus Y whose restriction to XX coincides with dXd_X and whose restriction to YY coincides with dYd_Y.✦ The ambient cost of an isometric amalgamation of two valid metric measure presentations XX and YY is the maximum of the Hausdorff distance between XX and YY and the Lévy-Prokhorov distance between the pushforward measures.✦ The Gromov-Hausdorff-Prokhorov distance dGHP(X,Y)d_{\mathrm{GHP}}(X, Y) between XX and YY is the infimum of ambient costs over all isometric amalgamations.✦ A Gromov-Hausdorff-Prokhorov test function FF is a bounded continuous real-valued function on valid metric measure presentations.✦ For any rooted planar map MM and scaling factor λ>0\lambda > 0, the rescaled metric measure presentation of MM is the presentation with vertex set V(M)V(M), distance λdM\lambda d_M, and measure μM\mu_M.✦

The Brownian continuum random tree is coded by a normalized Brownian excursion. Let B=(B1,B2,B3)B = (B_1, B_2, B_3) be three independent standard one-dimensional Brownian motions on a probability space (Ω,ℱ,ℙ)(\Omega, \mathcal{F}, \mathbb{P}). The normalized Brownian excursion e=(e(t))t∈[0,1]e = (e(t))_{t \in [0, 1]} is given by the three-dimensional Bessel bridge, e(t)=∑i=13(Bi(t)−tBi(1))2.e(t) = \sqrt{\sum_{i=1}^3 \bigl(B_i(t) - t B_i(1)\bigr)^2}.(3)✦ The excursion pseudometric ded_e on the unit interval [0,1][0, 1] is defined by de(s,t)=e(s)+e(t)−2minr∈[s∧t,s∨t]e(r).d_e(s, t) = e(s) + e(t) - 2 \min_{r \in [s \wedge t, s \vee t]} e(r).(4)✦ The Brownian continuum random tree (𝒯e,de,μe)(\mathcal{T}_e, d_e, \mu_e) is the quotient metric measure space [0,1]/{de=0}[0, 1] / \{d_e = 0\} endowed with the quotient metric ded_e and the pushforward measure μe\mu_e of the uniform Lebesgue measure on [0,1][0, 1].✦

The inventory bijection relates maps to trees through a burger and order alphabet. For q>4q > 4, the parameter transformation sets t=qt = \sqrt{q}, p=t/(t+2)∈(1/2,1)p = t / (t + 2) \in (1/2, 1), and u=(1−p)/16u = (1-p)/16. The alphabet contains burger letters and order letters, each of two types, together with a flexible order letter. Letters carry probability weights: burger letters have mass 1/41/4, whereas flexible and fixed orders have masses p/2p/2 and (1−p)/4(1-p)/4.✦ The inventory stack evaluation processes words freshest-first, and an order letter consumes the freshest eligible burger. An empty-reducing typed word ww is a typed word whose stack evaluation empties the stack.✦ A connected core cc is an empty-reducing typed word whose two-stack matching has a connected chord intersection graph.✦ A core tree TT is an ordered tree whose nodes are labelled by cores, with node arity equal to the length of the core word.✦ The inventory bijection establishes an equivalence between rooted planar maps M(x)M(x) of size nn and marked tree codes xx drawn under the law ℙdraw,n\mathbb{P}_{\mathrm{draw}, n}, corresponding bijectively to core trees of size 2(n+1)+12(n+1)+1.

The block generating function Φ(y)=1+∑m≥1bmym\Phi(y) = 1 + \sum_{m \ge 1} b_m y^m gives rise to a critical even offspring distribution on non-negative integers whose mass at 2m2m is the block mass bmb_m. This distribution has mean one, finite variance σ2>0\sigma^2 > 0, and strictly positive masses at zero and two. For the Galton-Watson tree with this offspring distribution conditioned on size 2n+12n+1, the spine ladder distribution governs the distribution of core marks along its exploration walk. The mean core metric increment β\beta is the expected graph distance in a block from its pole to a distinguished corner under the spine distribution.✦ The scaling constant c(q)c(q) of the continuum limit is defined by c(q)=σ(p)22β(p),c(q) = \frac{\sigma(p)}{2\sqrt{2}\,\beta(p)},(5)✦ and graph distances on maps of size nn are rescaled by c(q)/n+1c(q) / \sqrt{n+1}.✦ The law of marked code draw ℙdraw,n\mathbb{P}_{\mathrm{draw}, n} on marked tree code samples of size nn is the probability measure obtained from the conditioned tree code law of the critical offspring distribution equipped with the conditional core mark distribution at each arity.✦

The Claim

Theorem 1. Let q>4q > 4 be a real Potts parameter. Let FF be a Gromov-Hausdorff-Prokhorov test function. Then the scaling constant c(q)=σ(p)/(22β(p))c(q) = \sigma(p) / (2\sqrt{2}\,\beta(p)) satisfies c(q)>0.c(q) > 0.(6)✦ Furthermore, the finite Fortuin-Kasteleyn planar maps converge weakly to the Brownian continuum random tree, limn→∞∑M∈ℳn∑A⊆E(M)wq(M,A)Zn(q)F(V(M),c(q)n+1dM,μM)=𝔼[F(𝒯e)],\lim_{n \to \infty} \sum_{M \in \mathcal{M}_n} \sum_{A \subseteq E(M)} \frac{w_q(M, A)}{Z_n(q)}\, F\!\left( V(M), \frac{c(q)}{\sqrt{n+1}}\, d_M, \mu_M \right) = \mathbb{E}[F(\mathcal{T}_e)],(7)✦ and under the law ℙdraw,n\mathbb{P}_{\mathrm{draw}, n} of marked tree codes xx encoding maps M(x)M(x), the marked map test observables satisfy limn→∞∫F(V(M(x)),c(q)n+1dM(x),μM(x))dℙdraw,n(x)=𝔼[F(𝒯e)].\lim_{n \to \infty} \int F\!\left( V(M(x)), \frac{c(q)}{\sqrt{n+1}}\, d_{M(x)}, \mu_{M(x)} \right) d\mathbb{P}_{\mathrm{draw}, n}(x) = \mathbb{E}[F(\mathcal{T}_e)].(8)✦

In the paper, the result reads: for every fixed real q>4q > 4, there is a deterministic constant c(q)∈(0,∞)c(q) \in (0,\infty) such that, as n→∞n \to \infty through all positive integers, (V(Mn),c(q)n−1/2dn,μn)(V(M_n), c(q) n^{-1/2} d_n, \mu_n) converges in distribution to (𝒯e,de,μe)(\mathcal{T}_e, d_e, \mu_e) in the Gromov-Hausdorff-Prokhorov topology on compact metric probability spaces modulo measure-preserving isometry.

What Is Checked

The scaling limit (7) is stated over the Potts parameter q>4q > 4, the finite probability space of rooted planar maps with edge subsets weighted by wq(M,A)w_q(M, A) with partition function Zn(q)Z_n(q) given in (2), the rescaled metric measure presentations (V(M),c(q)(n+1)−1/2dM,μM)(V(M), c(q)(n+1)^{-1/2} d_M, \mu_M) equipped with normalized degree measure μM\mu_M, the continuum random tree (𝒯e,de,μe)(\mathcal{T}_e, d_e, \mu_e) generated by the three-dimensional Bessel bridge excursion ee in (3), and the class of bounded continuous test functions with respect to the Gromov-Hausdorff-Prokhorov distance on valid compact metric probability spaces.

Cyclic rotation by the good rotation of a discrete bridge yields a valid Lukasiewicz tree code✦, the sequence of core tree degrees forms a discrete excursion✦, and the planar map conditions are invariant under root-preserving dart relabelling.✦