OctoInformBeta

OpenAI preprint, September 2026 · §3. Ladder cuts and a coefficient lower bound · §3

Proposition 3.1 [Coefficient lower bound]: Quadratic partial-sum estimates for empty-reducing words and a (1−x)−1/2(1-x)^{-1/2} growth bound for x∈[3/4,1)x \in [3/4, 1)

Abel lower bound for Fortuin-Kasteleyn maps with q>4q > 4..

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

Fix a Potts parameter q>4q > 4, and weight each rooted planar map by its Fortuin-Kasteleyn (FK) loop partition function. How fast does the edge-weighted generating series of these map partition coefficients, evaluated at its inventory radius uu for x∈[3/4,1)x \in [3/4, 1), grow as xx approaches 11 from below?

These notes cover the definitions of rooted FK maps and the burger-order inventory model, together with the partial-sum and Abel lower bounds.

Rooted Planar Maps and FK Coefficients

For each nonnegative integer nn, the dart set DnD_n of a map with n+1n + 1 edges is {0,…,2(n+1)−1}\{0, \dots, 2(n + 1) - 1\}, indexing its 2(n+1)2(n + 1) directed edges.✦ For a permutation σ\sigma of DnD_n, the orbit [d]σ[d]_{\sigma} of a dart d∈Dnd \in D_n under σ\sigma is the equivalence class of darts belonging to the cycle of σ\sigma containing dd.✦ Given two permutations α\alpha and σ\sigma of DnD_n and a subset AA of the orbits of α\alpha, the spanning graph determined by α\alpha, σ\sigma, and AA is the simple graph on the orbits of σ\sigma in which two distinct orbits vv and ww of σ\sigma are adjacent whenever some dart d∈Dnd \in D_n with [d]α∈A[d]_{\alpha} \in A joins vv and ww, meaning either [d]σ=v[d]_{\sigma} = v and [α(d)]σ=w[\alpha(d)]_{\sigma} = w, or [d]σ=w[d]_{\sigma} = w and [α(d)]σ=v[\alpha(d)]_{\sigma} = v.✦

A pair (α,σ)(\alpha, \sigma) of permutations of DnD_n satisfies the planar map conditions if α\alpha is an involution without fixed points (α(α(d))=d\alpha(\alpha(d)) = d and α(d)≠d\alpha(d) \ne d for all d∈Dnd \in D_n), the spanning graph determined by α\alpha, σ\sigma, and all orbits of α\alpha is connected, and the number of orbits of σ\sigma plus the number of orbits of σ∘α\sigma \circ \alpha equals n+3n + 3 (the genus-zero condition).✦ A labelled planar map M=(α,σ)M = (\alpha, \sigma) with n+1n + 1 edges is a pair of permutations (α,σ)(\alpha, \sigma) of DnD_n satisfying the planar map conditions.✦ An isomorphism of rooted maps between two labelled planar maps M=(αM,σM)M = (\alpha_M, \sigma_M) and N=(αN,σN)N = (\alpha_N, \sigma_N) with n+1n + 1 edges is a permutation rr of DnD_n satisfying r(0)=0r(0) = 0 as well as αN(r(d))=r(αM(d))\alpha_N(r(d)) = r(\alpha_M(d)) and σN(r(d))=r(σM(d))\sigma_N(r(d)) = r(\sigma_M(d)) for every d∈Dnd \in D_n.✦ A rooted planar map MM with n+1n + 1 edges is an equivalence class of labelled planar maps with n+1n + 1 edges under the equivalence relation generated by isomorphisms of rooted maps, and the set of such rooted planar maps is ℳn\mathcal{M}_n.✦

For a rooted planar map M∈ℳnM \in \mathcal{M}_n, the dart reversal involution αM\alpha_M and vertex rotation permutation σM\sigma_M are the first and second permutations on DnD_n of a representative labelled map, respectively pairing opposite directed edges and cyclically ordering the darts around each vertex. The set of edges E(M)E(M) of a rooted planar map M∈ℳnM \in \mathcal{M}_n is the set of orbits of αM\alpha_M, each orbit being an edge of MM.✦ Given an involution α\alpha on DnD_n and a subset AA of the orbits of α\alpha, the partial dart reversal αA\alpha_A is the permutation of DnD_n that sends each dart d∈Dnd \in D_n with [d]α∈A[d]_{\alpha} \in A to α(d)\alpha(d) and fixes every dart d∈Dnd \in D_n with [d]α∉A[d]_{\alpha} \notin A.✦

For a permutation τ\tau of DnD_n, the number of orbits of τ\tau is denoted by ℓ(τ)\ell(\tau), and for a rooted planar map M∈ℳnM \in \mathcal{M}_n and a real parameter tt, the loop polynomial LM(t)L_M(t) is the real number LM(t)=∑A⊆E(M)tℓ(σM∘(αM)A)−1.L_M(t) = \sum_{A \subseteq E(M)} t^{\ell(\sigma_M \circ (\alpha_M)_A) - 1}.(1)✦ For a real parameter qq, the rooted FK map coefficients (an)n≥0(a_n)_{n \ge 0} are the sequence of real numbers defined by a0=1a_0 = 1 and, for every nonnegative integer nn, an+1=∑M∈ℳnLM(q).a_{n+1} = \sum_{M \in \mathcal{M}_n} L_M(\sqrt{q}).(2)✦

The Burger Inventory Model

An FK model parameter pp is a real probability in (1/2,1)(1/2, 1).✦ Given a Potts parameter q>4q > 4, the associated FK model parameter pp is p=qq+2.p = \frac{\sqrt{q}}{\sqrt{q} + 2}.(3)✦ The inventory radius uu for q>4q > 4 is (1−p)/16(1 - p)/16, with pp given by (3).✦

An inventory order is either a fixed order specifying one of two Boolean burger types or a flexible order.✦ A letter of the burger-order alphabet 𝒜\mathcal{A} is either a burger token of one of the two Boolean types or an order token carrying an inventory order.✦ An inventory stack evaluation state consists of a sequence of pending orders and a stack of unused burgers stored freshest-first.✦ The evaluation of a word of letters in 𝒜\mathcal{A} from an initial evaluation state ee returns ee when the word is empty, and on a word starting with letter aa followed by suffix w′w' updates ee by a single-step transition on aa (in which an order consumes the freshest eligible burger) and recursively evaluates w′w'.✦ The word reduction of a word ww is the evaluation state obtained by evaluating ww starting with empty order and burger lists.✦ A word ww of letters in 𝒜\mathcal{A} is empty-reducing if its word reduction yields the empty evaluation state, with neither pending orders nor remaining burgers.✦

For an FK model parameter p∈(1/2,1)p \in (1/2, 1), the single-letter probability distribution νp\nu_p on 𝒜\mathcal{A} is the probability distribution that assigns mass 1/41/4 to each of the two burger letters, p/2p/2 to the flexible order letter, and (1−p)/4(1 - p)/4 to each of the two fixed order letters.✦ On 𝒜ι\mathcal{A}^{\iota} for any index set ι\iota, the independent and identically distributed (i.i.d.) inventory stream measure ℙpι\mathbb{P}_p^{\iota} is the product measure of copies of νp\nu_p.✦ Given an infinite stream x=(x0,x1,…)∈𝒜ℕx = (x_0, x_1, \dots) \in \mathcal{A}^{\mathbb{N}} and a nonnegative integer NN, the prefix (x0,…,xN−1)(x_0, \dots, x_{N-1}) defines an initial word wN(x)w_N(x) of length NN, and for an FK model parameter p∈(1/2,1)p \in (1/2, 1) the empty-reducing word probability πN(p)\pi_N(p) is πN(p)=ℙpℕ(wN(x) is empty-reducing).\pi_N(p) = \mathbb{P}_p^{\mathbb{N}}\bigl(w_N(x) \text{ is empty-reducing}\bigr).(4)✦

Partial-sum and Abel Lower Bounds

At any FK model parameter p∈(1/2,1)p \in (1/2, 1), the empty-reducing word probabilities (4) satisfy a quadratic partial-sum lower bound: there exist δ>0\delta > 0 and a positive integer AA such that, for every nonnegative integer HH, δH≤∑N=02AH2(N+1)πN+2(p).\delta H \le \sum_{N=0}^{2AH^2} (N + 1) \pi_{N+2}(p).(5)✦ For each FK model parameter p∈(1/2,1)p \in (1/2, 1), there exist δ>0\delta > 0 and a positive integer AA such that, for every nonnegative integer HH, δH≤∑n=02AH2nπ2n(p).\delta H \le \sum_{n=0}^{2AH^2} n \pi_{2n}(p).(6)✦

Let (bn)n≥0(b_n)_{n \ge 0} be a sequence of nonnegative real numbers, let δ>0\delta > 0, and let AA be a nonnegative integer such that δH≤∑n=02AH2bn\delta H \le \sum_{n=0}^{2AH^2} b_n for every nonnegative integer HH. Then there exists c>0c > 0 such that, whenever x∈[3/4,1)x \in [3/4, 1) and ∑n=0∞bnxn\sum_{n=0}^{\infty} b_n x^n converges, c1−x≤∑n=0∞bnxn.\frac{c}{\sqrt{1-x}} \le \sum_{n=0}^{\infty} b_n x^n.(7)✦ When the coefficients are the even-length empty-reducing probabilities for an FK model parameter p∈(1/2,1)p \in (1/2, 1), there exists c>0c > 0 such that, for all real x∈[3/4,1)x \in [3/4, 1), c1−x≤∑n=0∞nπ2n(p)xn.\frac{c}{\sqrt{1-x}} \le \sum_{n=0}^{\infty} n \pi_{2n}(p) x^n.(8)✦

The Claim

Theorem 1. Let q>4q > 4 be a Potts parameter, (an)n≥0(a_n)_{n \ge 0} the corresponding rooted FK map coefficients, and uu the inventory radius. Then there exists c>0c > 0 such that, for every real x∈[3/4,1)x \in [3/4, 1), c1−x≤∑n=0∞nanunxn.\frac{c}{\sqrt{1-x}} \le \sum_{n=0}^{\infty} n a_n u^n x^n.(9)✦

In the paper, the result reads: there is c>0c > 0 such that, for all real x<1x < 1 sufficiently close to 11, ∑n≥1nanunxn≥c(1−x)−1/2\sum_{n \ge 1} n a_n u^n x^n \ge c (1 - x)^{-1/2}. The bound (9) replaces the qualitative proximity condition on x<1x < 1 with the explicit range 3/4≤x<13/4 \le x < 1 and begins the sum at n=0n = 0.

What Is Checked

The inequality (9) is stated over any Potts parameter q>4q > 4, the rooted FK map coefficients (2) defined from rooted planar maps via the loop polynomial (1), and the inventory radius uu determined by (3). The inventory bounds (5), (6), and (8) are stated over any FK model parameter p∈(1/2,1)p \in (1/2, 1) and the empty-reducing word probabilities (4) under the i.i.d. inventory stream measure, while (7) is stated over any nonnegative real sequence (bn)n≥0(b_n)_{n \ge 0} satisfying a quadratic partial-sum lower bound.

For these constructions, the library also proves that for any nonnegative integer nn, any involution α\alpha on DnD_n, and any subset AA of the orbits of α\alpha, the partial dart reversal αA\alpha_A is an involution.✦