OctoInformBeta

OpenAI preprint, September 2026 · §2. A graph of large distance classes · §2

Proposition 2.2 [The prepared graph]: Large distance classes of arbitrarily large cores form prepared graphs with errors θ/(j+2)→0\theta/(j+2)\to 0, each with a model in a number field containing −1\sqrt{-1}

Failure of the weak pinned distances statement gives prepared graphs.

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

A configuration is a finite set PP of points in the Euclidean plane ℝ2\mathbb{R}^2, with n=|P|n=|P|. The distance class of a point yy about a point xx is the set of points of PP other than xx at distance ∥y−x∥\|y-x\| from xx✦, and the pinned multiplicity kP(x,y)k_P(x,y) is its size✦. An ordered pair (x,y)(x,y) of distinct points of PP is rich at exponent ss when kP(x,y)≥nsk_P(x,y)\ge n^s.✦ The rich pair fraction of PP at exponent ss is the number of rich pairs divided by n(n−1)n(n-1)✦, and the extremal fraction Fn(s)F_n(s) is the supremum of the rich pair fraction over configurations of nn points when n≥2n\ge 2, and 00 when n<2n<2✦. The upper limit Θ\Theta of the extremal fraction and the critical exponent s*s_* are Θ(s)=limsupn→∞Fn(s),s*=sup⁡{s>0:Θ(s)>0}.\Theta(s)=\limsup_{n\to\infty}F_n(s),\qquad s_*=\sup\{s>0:\Theta(s)>0\}.(1) The weak pinned distances statement asserts that Fn(s)→0F_n(s)\to 0 as n→∞n\to\infty for every exponent s>0s>0.✦

Proposition 2.2 supposes that this statement fails. It then finds arbitrarily large subsets CC of configurations PP on which the distance classes of size at least nsn^s, read as directed edges from the point xx about which a class is taken to its points in CC, form graphs of edge density close to θ=Θ(s)>0\theta=\Theta(s)>0 and of large minimum degree, each with a model in a number field, that is, injective coordinates Z1,Z2Z_1,Z_2 into the field for which the product (Z1(y)−Z1(x))(Z2(y)−Z2(x))(Z_1(y)-Z_1(x))(Z_2(y)-Z_2(x)) is constant on each such distance class.

For maps Z1,Z2Z_1,Z_2 from a set into a field, write t(x,y)=(Z1(y)−Z1(x))(Z2(y)−Z2(x))t(x,y)=(Z_1(y)-Z_1(x))(Z_2(y)-Z_2(x)).

Fiber Systems

A fiber system on a finite vertex set VV is a finite family of fibers e∈Ie\in I, each with a source xe∈Vx_e\in V and a target set Fe⊂VF_e\subset V not containing xex_e, such that distinct fibers with the same source have disjoint target sets.✦ The edge set EE of a fiber system is the set of pairs (xe,y)(x_e,y) with e∈Ie\in I and y∈Fey\in F_e.✦ Each vertex xx has a total degree deg⁡(x)\deg(x), the number of edges leaving xx plus the number entering it. The in-neighborhood N−(y)N^-(y) of a vertex yy is the set of vertices xx with (x,y)∈E(x,y)\in E.✦

On a finite set BB, the uniform law gives mass 1/|B|1/|B| to each point of BB✦, and the uniform pair law gives mass 1/(|B|(|B|−1))1/(|B|(|B|-1)) to each ordered pair of distinct points of BB✦. For a fiber system on VV, α\alpha is the uniform law on VV and π\pi is the uniform pair law on VV, and π(D2)\pi(D^2) is the π\pi-mass of D×DD\times D for D⊂VD\subset V; μe\mu_e is the uniform law on FeF_e and ρy\rho_y the uniform law on N−(y)N^-(y). The edge law qq is the uniform law on EE, with q(x,y)=1/|E|q(x,y)=1/|E| on edges and q(x,y)=0q(x,y)=0 elsewhere.✦ A fiber ee has weight |Fe|/|E||F_e|/|E|✦, and a vertex yy has incoming weight |N−(y)|/|E||N^-(y)|/|E|✦.

The heavy count Hλ(D)H_\lambda(D) of a subset D⊂VD\subset V at threshold λ\lambda is the sum of |Fe∩D||F_e\cap D| over the fibers ee with source in DD that put at least the proportion λ\lambda of their targets in DD: Hλ(D)=∑e∈I:xe∈D,|Fe∩D|≥λ|Fe||Fe∩D|.H_\lambda(D)=\sum_{e\in I:\ x_e\in D,\ |F_e\cap D|\ge\lambda|F_e|}|F_e\cap D|.(2)✦

A fiber ee is active at level τ\tau in a subset C⊂VC\subset V when its source lies in CC and more than τ\tau of its targets lie in CC.✦ The edges at level τ\tau in CC are the pairs (xe,y)(x_e,y) with ee active and y∈Fe∩Cy\in F_e\cap C✦; they form the set Eτ(C)E_\tau(C)✦, and deg⁡τ,C(x)\deg_{\tau,C}(x) is the number of these edges leaving xx plus the number entering xx✦. The restriction of the system to CC at level τ\tau is the fiber system on CC whose fibers are the active fibers, each with source xex_e and target set Fe∩CF_e\cap C.✦

Prepared Graphs and Cores

For reals θ,γ,c,η\theta,\gamma,c,\eta, a prepared graph with parameters (θ,γ,c,η)(\theta,\gamma,c,\eta) is a fiber system on a vertex set VV with n=|V|≥2n=|V|\ge 2 in which every fiber has more than nγn^\gamma targets, (θ−η)n(n−1)≤|E|≤(θ+η8)n(n−1),deg⁡(x)≥2(θ−η)(n−1)(x∈V),(\theta-\eta)\,n(n-1)\le|E|\le\big(\theta+\tfrac{\eta}{8}\big)\,n(n-1),\qquad \deg(x)\ge 2(\theta-\eta)(n-1)\ \ (x\in V),(3) and every subset D⊂VD\subset V has H1/log⁡n(D)≤(θ+η)|D|(|D|−1),H1/log⁡n(D)≤η|D|(|D|−1)if |D|≤n1−c.H_{1/\log n}(D)\le(\theta+\eta)\,|D|(|D|-1),\qquad H_{1/\log n}(D)\le\eta\,|D|(|D|-1)\ \ \text{if } |D|\le n^{1-c}.(4)✦

The distance class system Φs(P)\Phi_s(P) of a configuration PP at exponent ss is the fiber system on PP whose fibers, the large distance classes, are the pairs (x,δ)(x,\delta) with δ=∥y−x∥\delta=\|y-x\| for some rich pair (x,y)(x,y); the fiber (x,δ)(x,\delta) has source xx and target set the points of PP at distance δ\delta from xx.✦ The core graph Gs(C)G_s(C) of a subset CC of a configuration PP of nn points is the restriction of Φs(P)\Phi_s(P) to CC at the cutoff level ns/log⁡nn^s/\log n.✦

For subsets C,D⊂PC,D\subset P, a level τ\tau and a threshold λ\lambda, a fiber ee of Φs(P)\Phi_s(P) is heavy in DD when it is active at level τ\tau in CC and |Fe∩D|≥λ|Fe∩C||F_e\cap D|\ge\lambda|F_e\cap C|, and the heavy count HλC(D)H^C_\lambda(D) is the sum of |Fe∩D||F_e\cap D| over the heavy fibers with source in DD✦; below, τ\tau is the cutoff level ns/log⁡nn^s/\log n.

A prepared core of PP with parameters (s,γ,c,η)(s,\gamma,c,\eta) is a subset C⊂PC\subset P of m≥2m\ge 2 points with m>(n+1)/2m>(n+1)/2 and mγ≤ns/log⁡nm^\gamma\le n^s/\log n such that, at the cutoff level τ=ns/log⁡n\tau=n^s/\log n and with edges and degrees taken in Φs(P)\Phi_s(P), (Θ(s)−η)m(m−1)≤|Eτ(C)|≤(Θ(s)+η8)m(m−1),deg⁡τ,C(x)≥2(Θ(s)−η)(m−1)(x∈C),(\Theta(s)-\eta)\,m(m-1)\le|E_\tau(C)|\le\big(\Theta(s)+\tfrac{\eta}{8}\big)\,m(m-1),\qquad \deg_{\tau,C}(x)\ge 2(\Theta(s)-\eta)(m-1)\ \ (x\in C),(5)✦ and every D⊂PD\subset P satisfies (4) with HCH^C, mm and Θ(s)\Theta(s) in place of HH, nn and θ\theta.

The Claim

Theorem 1 (Proposition 2.2).

  1. Suppose there is an exponent s0>0s_0>0 such that Fn(s0)F_n(s_0) does not tend to 00. Then there are exponents 0<c<γ<s<10<c<\gamma<s<1 with θ=Θ(s)>0\theta=\Theta(s)>0 and, for each j∈ℕj\in\mathbb{N}, a configuration PjP_j and a subset Cj⊂PjC_j\subset P_j with |Cj|>jandGs(Cj)a prepared graph with parameters(θ,γ,c,θj+2).|C_j|>j\quad\text{and}\quad G_s(C_j)\ \text{a prepared graph with parameters}\ \Big(\theta,\gamma,c,\frac{\theta}{j+2}\Big).(6)✦ These can be chosen so that also, for each jj, there are a number field KjK_j inside the algebraic closure ℚ¯\overline{\mathbb{Q}} of ℚ\mathbb{Q}, an element i∈Kji\in K_j with i2=−1i^2=-1, injective maps Z1,Z2:Cj→KjZ_1,Z_2\colon C_j\to K_j and a nonzero te∈Kjt_e\in K_j for each fiber ee of Gs(Cj)G_s(C_j), with t(xe,y)=tefor every fiber e of Gs(Cj) and every y∈Fe∩Cj.t(x_e,y)=t_e\quad\text{for every fiber $e$ of $G_s(C_j)$ and every } y\in F_e\cap C_j.(7)✦

  2. Under the hypothesis of (a), let γ,c,s\gamma,c,s be reals with 0<γ<s0<\gamma<s, 0<c<10<c<1, (1−c)s*<s(1-c)\,s_*<s and Θ\Theta continuous at ss. Then for every η>0\eta>0 and every M∈ℕM\in\mathbb{N} there are a configuration PP and a subset C⊂PC\subset P with |C|>MandCa prepared core of P with parameters(s,γ,c,η),|C|>M\quad\text{and}\quad C\ \text{a prepared core of $P$ with parameters}\ (s,\gamma,c,\eta),(8)✦ and there are a configuration PP and a subset C⊂PC\subset P with |C|>MandGs(C)a prepared graph with parameters(Θ(s),γ,c,η).|C|>M\quad\text{and}\quad G_s(C)\ \text{a prepared graph with parameters}\ (\Theta(s),\gamma,c,\eta).(9)✦

  3. Let PP be a configuration of n≥2n\ge 2 points, let s,r,θ,τs,r,\theta,\tau be reals and η>0\eta>0, and let m0m_0 be an integer with 2≤m0<n2\le m_0<n and m0(m0−1)≤n(n−1)/2m_0(m_0-1)\le n(n-1)/2. Suppose 0≤τ<ns0\le\tau<n^s, τ≤η8ns\tau\le\frac{\eta}{8}n^s, the rich pair fraction of PP at exponent ss is at least θ−η8\theta-\frac{\eta}{8}, and every integer ℓ\ell with m0≤ℓ≤nm_0\le\ell\le n has ℓr≤τ\ell^r\le\tau and Fℓ(r)≤θ+η8F_\ell(r)\le\theta+\frac{\eta}{8}. Then there is a subset C⊂PC\subset P of m>m0m>m_0 points with, in Φs(P)\Phi_s(P) at level τ\tau, (θ−η)m(m−1)≤|Eτ(C)|≤(θ+η8)m(m−1),deg⁡τ,C(x)≥2(θ−η)(m−1)(x∈C).(\theta-\eta)\,m(m-1)\le|E_\tau(C)|\le\big(\theta+\tfrac{\eta}{8}\big)\,m(m-1),\qquad \deg_{\tau,C}(x)\ge 2(\theta-\eta)(m-1)\ \ (x\in C).(10)✦

  4. Let a fiber system on VV, with n=|V|n=|V|, be a prepared graph with parameters (θ,γ,c,η)(\theta,\gamma,c,\eta), where η<θ\eta<\theta. Then q(x,y)≤1θ−ηπ(x,y)(x,y∈V).q(x,y)\le\frac{1}{\theta-\eta}\,\pi(x,y)\qquad(x,y\in V).(11)✦ If also η≥0\eta\ge 0, then every vertex xx and every subset D⊂VD\subset V satisfy ∑y∈Vq(x,y)+∑y∈Vq(y,x)≥2θ−ηθ+η/8α(x),\sum_{y\in V}q(x,y)+\sum_{y\in V}q(y,x)\ge 2\,\frac{\theta-\eta}{\theta+\eta/8}\,\alpha(x),(12)✦ H1/log⁡n(D)|E|≤θ+ηθ−ηπ(D2),\frac{H_{1/\log n}(D)}{|E|}\le\frac{\theta+\eta}{\theta-\eta}\,\pi(D^2),(13)✦ and, when |D|≤n1−c|D|\le n^{1-c}, H1/log⁡n(D)|E|≤ηθ−ηπ(D2).\frac{H_{1/\log n}(D)}{|E|}\le\frac{\eta}{\theta-\eta}\,\pi(D^2).(14)✦

  5. Let a fiber system on VV with |V|≥2|V|\ge 2 satisfy q≤Bπq\le B\,\pi pointwise for a real BB. Then every vertex xx satisfies ∑e∈I|Fe||E|μe(x)≤Bα(x),\sum_{e\in I}\frac{|F_e|}{|E|}\,\mu_e(x)\le B\,\alpha(x),(15)✦ ∑y∈V|N−(y)||E|ρy(x)≤Bα(x).\sum_{y\in V}\frac{|N^-(y)|}{|E|}\,\rho_y(x)\le B\,\alpha(x).(16)✦ Every fiber system on VV and every real T≥0T\ge 0 satisfy ∑y∈V:|N−(y)|<T|N−(y)||E|≤|V|T|E|.\sum_{y\in V:\ |N^-(y)|<T}\frac{|N^-(y)|}{|E|}\le\frac{|V|\,T}{|E|}.(17)✦

  6. Let PP be a configuration, ss a real and C⊂PC\subset P a subset such that every fiber of Gs(C)G_s(C) has a target. Then there are a number field K⊂ℚ¯K\subset\overline{\mathbb{Q}}, an element i∈Ki\in K with i2=−1i^2=-1, injective maps Z1,Z2:C→KZ_1,Z_2\colon C\to K and a nonzero te∈Kt_e\in K for each fiber ee of Gs(C)G_s(C) such that, for all a1,a2,a3,a4∈Ca_1,a_2,a_3,a_4\in C, t(xe,y)=te(y∈Fe∩C),t(a1,a2)=t(a3,a4)⇔∥a1−a2∥=∥a3−a4∥.t(x_e,y)=t_e\ \ (y\in F_e\cap C),\qquad t(a_1,a_2)=t(a_3,a_4)\iff\|a_1-a_2\|=\|a_3-a_4\|.(18)✦

  7. Let (xa)a∈A(x_a)_{a\in A} be a finite family of distinct points of the plane. Then there are a number field K⊂ℚ¯K\subset\overline{\mathbb{Q}}, an element i∈Ki\in K with i2=−1i^2=-1 and injective maps Z1,Z2:A→KZ_1,Z_2\colon A\to K such that, for all a1,a2,a3,a4∈Aa_1,a_2,a_3,a_4\in A, t(a1,a2)=t(a3,a4)⇔∥xa1−xa2∥=∥xa3−xa4∥.t(a_1,a_2)=t(a_3,a_4)\iff\|x_{a_1}-x_{a_2}\|=\|x_{a_3}-x_{a_4}\|.(19)✦

  8. Let 𝒮\mathcal{S} be a set and 𝒯\mathcal{T} a finite set of polynomials with rational coefficients in finitely many variables, and suppose some complex point zz has p(z)=0p(z)=0 for p∈𝒮p\in\mathcal{S} and p(z)≠0p(z)\neq 0 for p∈𝒯p\in\mathcal{T}. Then there are a number field K⊂ℚ¯K\subset\overline{\mathbb{Q}} and a point ww with coordinates in KK such that p(w)=0(p∈𝒮),p(w)≠0(p∈𝒯).p(w)=0\ \ (p\in\mathcal{S}),\qquad p(w)\neq 0\ \ (p\in\mathcal{T}).(20)✦

In the paper, the result reads: if Theorem 1.1, the weak pinned distances statement, fails, there are finite sets P⊂ℝ2P\subset\mathbb{R}^2 of real algebraic points, with n=|P|→∞n=|P|\to\infty, and nonempty directed graphs EE on PP with densities dn→θd_n\to\theta and every fiber of size |Fe|≥ns−o(1)|F_e|\ge n^{s-o(1)} such that, for a fixed B0B_0, q≤B0πq\le B_0\pi, qX,qY≤B0αq_X,q_Y\le B_0\alpha and (qX+qY)/2≥(1−o(1))α(q_X+q_Y)/2\ge(1-o(1))\alpha pointwise (2.2), qY{y:|N−(y)|<nγ}=O(nγ−1)q_Y\{y:|N^-(y)|<n^\gamma\}= O(n^{\gamma-1}) (2.3), and, with λn=1/log⁡n\lambda_n=1/\log n, every D⊂PD\subset P with m=|D|m=|D| and pe=μe(D)p_e=\mu_e(D) has 𝔼e1D(xe)pe1{pe≥λn}≤(1+o(1))π(D2)\mathbb{E}_e\,1_D(x_e)\,p_e\,1\{p_e\ge\lambda_n\}\le (1+o(1))\,\pi(D^2) if m>n1−cm>n^{1-c} and ≤o(1)π(D2)\le o(1)\,\pi(D^2) if m≤n1−cm\le n^{1-c} (2.4), all errors uniform, where qXq_X and qYq_Y are the marginals of qq and 𝔼e\mathbb{E}_e is the expectation over fibers.

The paper’s bounds (2.2) to (2.4) appear as parts (d) and (e).

What Is Checked

The claim is stated over configurations in the Euclidean plane ℝ2\mathbb{R}^2, the extremal fraction Fn(s)F_n(s) of rich ordered pairs with its upper limit Θ\Theta and the critical exponent s*s_*, fiber systems with their edge law, uniform laws and heavy counts, the distance class system Φs(P)\Phi_s(P) with its core graphs, prepared graphs and prepared cores, and number fields inside ℚ¯\overline{\mathbb{Q}}.

Beyond the claim, the library shows that the large distance classes of a configuration satisfy the two conditions of a fiber system, and that the restriction of a fiber system to a subset at a level is again a fiber system.