OctoInformBeta

OpenAI preprint, September 2026 · §7. Bounded scale and a limiting fiber map · §7

§7 (proof): the bounded case is impossible, and Theorem 1.1: One archimedean place, nonatomic weak limits on the Riemann sphere ℂˆ\hat{\mathbb{C}} and no limiting fiber configuration when the scale of the models stays bounded

Prepared arithmetic models cannot grow with vanishing error.

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

Write a point of the plane as u+ibu+ib and put Z1=u+ibZ_1=u+ib, Z2=u−ibZ_2=u-ib. The product t(x,y)=(Z1(y)−Z1(x))(Z2(y)−Z2(x))t(x,y)=\bigl(Z_1(y)-Z_1(x)\bigr)\bigl(Z_2(y)-Z_2(x)\bigr)(1) is then the squared distance |y−x|2|y-x|^2, so the points of a configuration on one circle about a pin xx are those with one value of t(x,⋅)t(x,\cdot). An arithmetic model keeps (1) and replaces the plane by a number field KK: the coordinates become injective maps Z1Z_1, Z2Z_2 of a finite vertex set into KK.

The question is whether prepared arithmetic models, arithmetic models of this kind whose grouping into fibers satisfies conditions on degrees and edge counts up to an error η\eta, can form a sequence whose number of vertices tends to infinity while η\eta tends to 00. The case to exclude first is the one in which the scale W2+AW_2+A of the models, the sum of the overlap scale and the potential scale, built from cell masses and potentials of both coordinates over all places of KK, stays bounded; there the statements pass from the number field to one archimedean place and to laws on the Riemann sphere.

Fiber Systems and Prepared Models

A fiber system GG on a finite vertex set VV, with fibers indexed by a finite set II, assigns to each fiber ee a source xe∈Vx_e\in V and a target set Fe⊆VF_e\subseteq V not containing xex_e, two distinct fibers with the same source having disjoint target sets.✦ Its edges are the pairs (xe,y)(x_e,y) with y∈Fey\in F_e✦; |E||E| is their number✦, qq the uniform law on them✦, μe\mu_e the uniform law on FeF_e, and we=|Fe|/|E|w_e=|F_e|/|E| the fiber weight, the qq-mass of the edges of ee✦.

The incoming neighbors of a vertex yy form the set N−(y)N^-(y) of vertices xx with (x,y)(x,y) an edge, and ρy\rho_y is the uniform law on N−(y)N^-(y). For a threshold T>0T>0, a vertex yy is retained when |N−(y)|≥T|N^-(y)|\ge T, and the omitted mass is the qq-mass of the edges ending at the other vertices: Ω(T)=∑y:|N−(y)|<T|N−(y)||E|.\Omega(T)=\sum_{y\,:\,|N^-(y)|<T}\frac{|N^-(y)|}{|E|}.(2)✦

For a proportion λ\lambda and a set D⊆VD\subseteq V, the heavy count Hλ(D)H_\lambda(D) is the sum of |Fe∩D||F_e\cap D| over the fibers ee with source in DD and |Fe∩D|≥λ|Fe||F_e\cap D|\ge\lambda|F_e|.✦ A fiber system GG on VV is a prepared graph with density θ\theta, exponents γ\gamma, cc and error η\eta when |V|≥2|V|\ge2, every vertex has degree, its number of out-neighbors plus its number of in-neighbors, at least 2(θ−η)(|V|−1)2(\theta-\eta)(|V|-1), every fiber has more than |V|γ|V|^\gamma targets, and, with λ=1/log⁡|V|\lambda=1/\log|V|, (θ−η)|V|(|V|−1)≤|E|≤(θ+η/8)|V|(|V|−1),Hλ(D)≤(θ+η)|D|(|D|−1)for every D⊆V,Hλ(D)≤η|D|(|D|−1)when |D|≤|V|1−c.\begin{gathered} (\theta-\eta)|V|(|V|-1)\le|E|\le(\theta+\eta/8)|V|(|V|-1),\\ H_\lambda(D)\le(\theta+\eta)|D|(|D|-1)\ \text{for every }D\subseteq V,\qquad H_\lambda(D)\le\eta|D|(|D|-1)\ \text{when }|D|\le|V|^{1-c}. \end{gathered}(3)✦

A prepared arithmetic model (G,Z1,Z2)(G,Z_1,Z_2) is a prepared graph GG on VV with injective maps Z1,Z2:V→KZ_1,Z_2\colon V\to K into a number field KK such that, for each fiber ee, the product t(xe,y)t(x_e,y) of (1) takes one value for all y∈Fey\in F_e.

Depths, Cell Masses and Scales

An ultrametric depth dd on a finite set VV is a symmetric real function of pairs of points with d(x,z)≥min⁡(d(x,y),d(y,z))d(x,z)\ge\min(d(x,y),d(y,z)) for pairwise distinct xx, yy, zz.✦ For an ultrametric depth dd on VV, a real number rr and a point x∈Vx\in V, the cell Cr(x)C_r(x) of dd at level rr is the set consisting of xx and the points yy with d(x,y)≥rd(x,y)\ge r.✦ A cell at level rr holding more than half of the points of VV is a giant✦, and the transformed family 𝒯r\mathcal T_r is the multiset of the cells at level rr that are not giants and of the complements of the giants, CC occurring τr(C)\tau_r(C) times✦. The cell mass of a set C⊆VC\subseteq V with at least two points, 00 for smaller sets, and the potential of a point xx are md(C)=∫ℝτr(C)dr,Sd(x)=∫ℝ𝟏[there is a giant at level r and none contains x]dr.m_d(C)=\int_{\mathbb R}\tau_r(C)\,dr,\qquad S_d(x)=\int_{\mathbb R}\mathbf 1\bigl[\text{there is a giant at level }r\text{ and none contains }x\bigr]\,dr.(4)

The grids come from a phase and grid sample ω\omega: a scale phase ξ\xi uniform on (0,log⁡2](0,\log2], a rotation angle ψ\psi uniform on (0,2π](0,2\pi], and a random shifted dyadic grid of the plane whose two axes each carry a shift uniform on (0,1](0,1] and a fair coarsening bit at every dyadic scale above 11. All of these are independent under the full sample law, whose mean is written 𝔼ω\mathbb E_\omega. The hierarchy depth of complex numbers zz, ww is δω(z,w)=−klog⁡2−ξ,\delta_\omega(z,w)=-k\log2-\xi,(5)✦ where 2k2^k is the side of the smallest grid cell containing e−ξ+iψze^{-\xi+i\psi}z and e−ξ+iψwe^{-\xi+i\psi}w.

Let Z:V→KZ\colon V\to K be injective. At an archimedean place vv of KK, with complex embedding ιv\iota_v, the sample ω\omega is good when δω\delta_\omega is ultrametric on the points ιvZ(x)\iota_vZ(x)✦, and the archimedean depth dv,ωd^{v,\omega} is δω(ιvZ(x),ιvZ(y))\delta_\omega(\iota_vZ(x),\iota_vZ(y)) at distinct points for a good sample and 00 otherwise✦. At a finite place vv the depth dv(x,y)=−log⁡|Z(x)−Z(y)|vd^v(x,y)=-\log|Z(x)-Z(y)|_v is the raw depth✦, and vv is bad when some difference Z(x)−Z(y)Z(x)-Z(y) of distinct points is not a unit at vv; there are finitely many bad places✦. With the archimedean weight σv=mult⁡(v)/[K:ℚ]\sigma_v=\operatorname{mult}(v)/[K:\mathbb Q], where mult⁡(v)\operatorname{mult}(v) is the multiplicity of the archimedean place vv, the global cell mass of ZZ is mZ(C)=∑v∣∞σv𝔼ωmdv,ω(C)+1[K:ℚ]∑vbadmdv(C),m_Z(C)=\sum_{v\mid\infty}\sigma_v\,\mathbb E_\omega\,m_{d^{v,\omega}}(C) +\frac{1}{[K:\mathbb Q]}\sum_{v\ \mathrm{bad}}m_{d^v}(C),(6) and the global potential SZ(x)S_Z(x) of ZZ at a point x∈Vx\in V is obtained from the sum in (6) by replacing each cell mass with the potential Sdv,ω(x)S_{d^{v,\omega}}(x) or Sdv(x)S_{d^v}(x).

For injective coordinates Z1Z_1, Z2Z_2 of VV, the total cell mass is m*=mZ1+mZ2m^*=m_{Z_1}+m_{Z_2}✦ and the total potential is S*=SZ1+SZ2S^*=S_{Z_1}+S_{Z_2}✦. With π\pi the uniform law on distinct ordered pairs of VV and α\alpha the uniform law on VV, the total of a function ff of subsets of VV, the overlap scale and the potential scale are I(f)=∑C⊆Vm*(C)f(C),W2=I(C↦π(C×C)),A=𝔼αS*.I(f)=\sum_{C\subseteq V}m^*(C)f(C),\qquad W_2=I\bigl(C\mapsto\pi(C\times C)\bigr),\qquad A=\mathbb E_\alpha S^*.(7) At one archimedean place vv, with d1d_1, d2d_2 the archimedean depths of Z1Z_1, Z2Z_2 for the sample ω\omega, the corresponding quantities are Iv(f)=𝔼ω∑C⊆V(md1(C)+md2(C))f(C),W(v)=Iv(C↦π(C×C))+𝔼ω𝔼α(Sd1+Sd2).I_v(f)=\mathbb E_\omega\sum_{C\subseteq V}\bigl(m_{d_1}(C)+m_{d_2}(C)\bigr)f(C),\qquad W(v)=I_v\bigl(C\mapsto\pi(C\times C)\bigr)+\mathbb E_\omega\mathbb E_\alpha(S_{d_1}+S_{d_2}).(8) The scale of the models at the place vv is W(v)W(v): the mean over ω\omega of the overlap term, the sum over cells of the cell masses of the two coordinates times the probability that a uniform ordered pair of distinct points lies in the cell, plus the potential term, the mean over a uniform point of the sum of the two potentials.

Laws on the Riemann Sphere

The Riemann sphere ℂˆ=ℂ∪{∞}\hat{\mathbb C}=\mathbb C\cup\{\infty\} is the one-point compactification of ℂ\mathbb C.✦ A law on ℂˆ\hat{\mathbb C} or on a power of it is a Borel probability measure, and complex values are read as points of ℂˆ\hat{\mathbb C}. A law β\beta is nonatomic when every point has β\beta-measure 00. For weights wi≥0w_i\ge0 of sum 11 on a finite set and a map UU from that set to a space XX, the finite law on XX is ∑iwiδU(i)\sum_iw_i\delta_{U(i)}✦, where δp\delta_p is the point mass at pp; the uniform law U*αU_*\alpha of a map U:V→XU\colon V\to X is the finite law of UU with all weights 1/|V|1/|V|.

Draw a fiber ee of a fiber system GG with probability wew_e and targets yy, zz, z′z' independently from μe\mu_e. For maps U1,U2:V→ℂU_1,U_2\colon V\to\mathbb C, the edge measure QQ on ℂˆ4\hat{\mathbb C}^4 and the expanded measure RR on ℂˆ4×ℂˆ2×ℂˆ2\hat{\mathbb C}^4\times\hat{\mathbb C}^2\times\hat{\mathbb C}^2 are the finite laws of (U1(xe),U2(xe),U1(y),U2(y))and(U1(xe),U2(xe),U1(y),U2(y);U1(z),U2(z);U1(z′),U2(z′)),\bigl(U_1(x_e),U_2(x_e),U_1(y),U_2(y)\bigr)\quad\text{and}\quad \bigl(U_1(x_e),U_2(x_e),U_1(y),U_2(y);\,U_1(z),U_2(z);\,U_1(z'),U_2(z')\bigr),(9)

For a function gg on VV, the fiber discrepancy and, for a threshold TT, the retained discrepancy compare with α\alpha the fiber laws and the laws of the incoming neighbors of retained vertices: DF(g)=∑ewe|𝔼μeg−𝔼αg|,DT(g)=∑y:|N−(y)|≥T|N−(y)||E||𝔼ρyg−𝔼αg|.D_F(g)=\sum_{e}w_e\bigl|\mathbb E_{\mu_e}g-\mathbb E_\alpha g\bigr|,\qquad D_T(g)=\sum_{y\,:\,|N^-(y)|\ge T}\frac{|N^-(y)|}{|E|} \bigl|\mathbb E_{\rho_y}g-\mathbb E_\alpha g\bigr|.(10)

For an injective map u:V→ℂu\colon V\to\mathbb C, the logarithmic cost of distinct vertices is ℓu(x,y)=|log⁡|u(x)−u(y)||\ell_u(x,y)=\bigl|\log|u(x)-u(y)|\bigr|, and 00 when x=yx=y✦; its mean over distinct ordered pairs is Λ(u)=𝔼πℓu\Lambda(u)=\mathbb E_\pi\ell_u✦. The map normalized at a base point x0x_0 is u‾(y)=u(y)−u(x0)\bar u(y)=u(y)-u(x_0).

The Claim

Theorem 1 (Selection of one archimedean place). Let Z1,j,Z2,j:Vj→KjZ_{1,j},Z_{2,j}\colon V_j\to K_j be injective maps of finite sets with at least two points into number fields, let gj,hj≥0g_j,h_j\ge0 be functions of subsets of VjV_j, and let C≥0C\ge0. Suppose that W2+A≤CW_2+A\le C for every jj, and that I(gj)→0I(g_j)\to0 and I(hj)→0I(h_j)\to0. Then there are archimedean places vjv_j of KjK_j with W(vj)≤2C+1for every j,Ivj(gj)→0,Ivj(hj)→0.W(v_j)\le2C+1\ \text{for every }j,\qquad I_{v_j}(g_j)\to0,\qquad I_{v_j}(h_j)\to0.(11)✦

Theorem 2 (Nonatomic weak limits). (a) Let νj\nu_j be laws on ℂˆ\hat{\mathbb C} converging weakly to β\beta. Suppose that for every ε>0\varepsilon>0 and z∈ℂz\in\mathbb C there are RR, depending on ε\varepsilon alone, and r>0r>0 such that, for all large jj, νj(ℂˆ\D‾(0,R))≤ε,νj(D(z,r))≤ε.\nu_j\bigl(\hat{\mathbb C}\setminus\bar D(0,R)\bigr)\le\varepsilon,\qquad \nu_j\bigl(D(z,r)\bigr)\le\varepsilon.(12)✦ Then β\beta is nonatomic.✦

(b) Let uj:Vj→ℂu_j\colon V_j\to\mathbb C be injective maps of finite sets with at least two points and |Vj|→∞|V_j|\to\infty, with base points xj∈Vjx_j\in V_j and a constant C≥0C\ge0 such that Λ(uj)≤C,∑y∈Vjℓuj(xj,y)≤C|Vj|for every j.\Lambda(u_j)\le C,\qquad\sum_{y\in V_j}\ell_{u_j}(x_j,y)\le C|V_j|\qquad\text{for every }j.(13)✦ If the uniform laws of the normalized maps u‾j\bar u_j converge weakly to β\beta, then β\beta is nonatomic.✦

Theorem 3 (No limiting fiber configuration). (a) Let GjG_j be fiber systems on finite sets VjV_j with at least two vertices, at least one edge and nonempty target sets, and let U1,j,U2,j:Vj→ℂU_{1,j},U_{2,j}\colon V_j\to\mathbb C be maps, with edge measures QjQ_j, expanded measures RjR_j and omitted masses Ωj\Omega_j. Suppose that for a constant M≥0M\ge0 and every jj, qj(x,y)≤Mπj(x,y)q_j(x,y)\le M\pi_j(x,y) for all xx, yy, and the product (U1,j(y)−U1,j(xe))(U2,j(y)−U2,j(xe))(U_{1,j}(y)-U_{1,j}(x_e))(U_{2,j}(y)-U_{2,j}(x_e)) takes one value on each target set FeF_e. Then there are no nonatomic laws β1\beta_1, β2\beta_2 on ℂˆ\hat{\mathbb C}, laws QQ on ℂˆ4\hat{\mathbb C}^4 and RR on ℂˆ4×ℂˆ2×ℂˆ2\hat{\mathbb C}^4\times\hat{\mathbb C}^2\times\hat{\mathbb C}^2 and thresholds Tj>0T_j>0 such that, weakly for the laws and for every continuous f:ℂˆ→ℝf\colon\hat{\mathbb C}\to\mathbb R, (U1,j)*αj→β1,(U2,j)*αj→β2,Qj→Q,Rj→R,DF(f∘U1,j)→0,DF(f∘U2,j)→0,DTj(f∘U1,j)→0,Ωj(Tj)→0.\begin{gathered} (U_{1,j})_*\alpha_j\to\beta_1,\quad(U_{2,j})_*\alpha_j\to\beta_2,\quad Q_j\to Q,\quad R_j\to R,\\ D_F(f\circ U_{1,j})\to0,\quad D_F(f\circ U_{2,j})\to0,\quad D_{T_j}(f\circ U_{1,j})\to0,\quad \Omega_j(T_j)\to0. \end{gathered}(14)✦

(b) Let GjG_j be as in (a) with moreover |Vj|→∞|V_j|\to\infty, and let u1,j,u2,j:Vj→ℂu_{1,j},u_{2,j}\colon V_j\to\mathbb C be injective maps satisfying the two conditions of (a) in place of U1,jU_{1,j}, U2,jU_{2,j}. Then there are no base points x1,jx_{1,j}, x2,jx_{2,j}, constant C≥0C\ge0 and thresholds Tj>0T_j>0 such that u1,ju_{1,j} at x1,jx_{1,j} and u2,ju_{2,j} at x2,jx_{2,j} satisfy (13) for every jj and, with the normalized maps u‾1,j\bar u_{1,j}, u‾2,j\bar u_{2,j} and for every continuous f:ℂˆ→ℝf\colon\hat{\mathbb C}\to\mathbb R, DF(f∘u‾1,j)→0,DF(f∘u‾2,j)→0,DTj(f∘u‾1,j)→0,Ωj(Tj)→0.D_F(f\circ\bar u_{1,j})\to0,\quad D_F(f\circ\bar u_{2,j})\to0,\quad D_{T_j}(f\circ\bar u_{1,j})\to0,\quad\Omega_j(T_j)\to0.(15)✦

Theorem 4 (Bounded subsequence). Let (aj)(a_j) be a real sequence that does not tend to +∞+\infty, and let pp be a property of indices that holds for all large jj. Then there are indices φ(n)→∞\varphi(n)\to\infty and a constant C≥0C\ge0 with aφ(n)≤Candp(φ(n))for every n.a_{\varphi(n)}\le C\quad\text{and}\quad p(\varphi(n))\qquad\text{for every }n.(16)✦

Theorem 5 (No prepared arithmetic sequence). Let θ>0\theta>0 and 0<c<γ<10<c<\gamma<1, and consider sequences of prepared arithmetic models (Gj,Z1,j,Z2,j)(G_j,Z_{1,j},Z_{2,j}) on vertex sets VjV_j with density θ\theta, exponents γ\gamma, cc and errors ηj\eta_j, such that |Vj|→∞,ηj→0,0≤ηj≤θ/2.|V_j|\to\infty,\qquad\eta_j\to0,\qquad0\le\eta_j\le\theta/2.(17)✦

(a) No such sequence has a constant C≥0C\ge0 with W2+A≤Cfor every j.W_2+A\le C\qquad\text{for every }j.(18)✦

(b) No such sequence exists.✦

In the paper, the result reads: the selection of one complex embedding, the nonatomic limits of the normalized laws and the limiting fiber map, together with §6, exclude the graph sequence of Proposition 2.2. Applied to the models of the paper’s Proposition 2.2, the nonexistence of sequences with (17) gives the paper’s Theorem 1.1, the weak pinned planar distance theorem.

What Is Checked

The claims are stated over fiber systems and prepared graphs (3), with their uniform edge law, fiber weights, incoming neighbors and omitted mass (2); over injective coordinates in number fields, with the ultrametric depths at their archimedean and finite places built from the random grids, and the totals and scales (7) and (8); and over Borel probability measures on ℂˆ\hat{\mathbb C} and its powers, with the uniform laws, the edge and expanded measures (9), the discrepancies (10) and the logarithmic costs. About these constructions the library proves that a finite law with nonnegative weights of sum 11, the uniform law of a map on a nonempty finite set among them, is a probability measure, and that the weights of the edge and expanded measures are nonnegative. The weights of the edge and expanded measures also sum to 11, so that both are probability measures. The library also proves that the archimedean depth at a place, for every sample, and the depth at a finite place are ultrametric depths: they are symmetric and satisfy the ultrametric inequality.