Prove2Me
Navigate
DiscoverCollectionsFormalpediaGalaxyBlogsUsersMomentumMy Missions+
Prove2Me
Log in
← Formalpedia

Conditional source and reverse direction bounds at the middle scale

Open
FalconerPath.middle_scale_preparation

by dangnguyen · Nov 5, 2026 · Mathlib 0df444a (Lean v4.33.1)

conditional-measuresfalconermiddle-scale

Let k∈Nk\in\mathbb Nk∈N and s,q,ζ,L,a,C∈Rs,q,\zeta,L,a,C\in\mathbb Rs,q,ζ,L,a,C∈R. For every j∈Nj\in\mathbb Nj∈N, let XjX_jXj​ consist of Nj∈NN_j\in\mathbb NNj​∈N, probabilities νj,ρj\nu_j,\rho_jνj​,ρj​ on Rk+1\mathbb R^{k+1}Rk+1, a translation vjv_jvj​, a real function mjm_jmj​, an error ej≥0e_j\ge0ej​≥0, and a Borel relation GjG_jGj​. Assume L,a>0L,a>0L,a>0, both supports lie in B‾(0,L)\overline B(0,L)B(0,L), and ∥x−z∥≥a\|x-z\|\ge a∥x−z∥≥a for every x∈supp⁡νjx\in\operatorname{supp}\nu_jx∈suppνj​ and z∈supp⁡ρjz\in\operatorname{supp}\rho_jz∈suppρj​. The measure ρj\rho_jρj​ is carried by the half-open unit grid cell Q(vj,0,0)Q(v_j,0,0)Q(vj​,0,0). On [0,1][0,1][0,1], mjm_jmj​ is nondecreasing and (k+1)(k+1)(k+1)-Lipschitz, mj(0)=0m_j(0)=0mj​(0)=0, and mj(u)≥su−ejm_j(u)\ge su-e_jmj​(u)≥su−ej​. Put Rj=2NjR_j=2^{N_j}Rj​=2Nj​. For every ℓ≤Nj\ell\le N_jℓ≤Nj​ and cell Q=Q(vj,ℓ,i)Q=Q(v_j,\ell,i)Q=Q(vj​,ℓ,i) with positive mass, require Rj−mj(ℓ/Nj)−ej≤ρj(Q)≤Rj−mj(ℓ/Nj)+ejR_j^{-m_j(\ell/N_j)-e_j}\le\rho_j(Q)\le R_j^{-m_j(\ell/N_j)+e_j}Rj−mj​(ℓ/Nj​)−ej​​≤ρj​(Q)≤Rj−mj​(ℓ/Nj​)+ej​​. For every 0≤ℓ≤h≤Nj0\le\ell\le h\le N_j0≤ℓ≤h≤Nj​ and nested cells Q′=Q(vj,h,i′)⊆Q=Q(vj,ℓ,i)Q'=Q(v_j,h,i')\subseteq Q=Q(v_j,\ell,i)Q′=Q(vj​,h,i′)⊆Q=Q(vj​,ℓ,i) with ρj(Q′)>0\rho_j(Q')>0ρj​(Q′)>0, require Rj−(mj(h/Nj)−mj(ℓ/Nj))−ej≤ρj(Q′)/ρj(Q)≤Rj−(mj(h/Nj)−mj(ℓ/Nj))+ejR_j^{-(m_j(h/N_j)-m_j(\ell/N_j))-e_j}\le\rho_j(Q')/\rho_j(Q)\le R_j^{-(m_j(h/N_j)-m_j(\ell/N_j))+e_j}Rj−(mj​(h/Nj​)−mj​(ℓ/Nj​))−ej​​≤ρj​(Q′)/ρj​(Q)≤Rj−(mj​(h/Nj​)−mj​(ℓ/Nj​))+ej​​. At Nj=0N_j=0Nj​=0 real division uses 0/0=00/0=00/0=0. Assume Gj⊆supp⁡νj×supp⁡ρjG_j\subseteq\operatorname{supp}\nu_j\times\operatorname{supp}\rho_jGj​⊆suppνj​×suppρj​. Define λj,z=(πz)#(νj∣{x:(x,z)∈Gj})\lambda_{j,z}=(\pi_z)_\#(\nu_j|_{\{x:(x,z)\in G_j\}})λj,z​=(πz​)#​(νj​∣{x:(x,z)∈Gj​}​), where πz(x)=∥x−z∥−1(x−z)\pi_z(x)=\|x-z\|^{-1}(x-z)πz​(x)=∥x−z∥−1(x−z) with value zero at x=zx=zx=z. For ρj\rho_jρj​-almost every zzz, every w∈Skw\in S^kw∈Sk and every 2−Nj≤u≤12^{-N_j}\le u\le12−Nj​≤u≤1 with u>0u>0u>0 satisfy λj,z(B(w,u))≤Rjejuq\lambda_{j,z}(B(w,u))\le R_j^{e_j}u^qλj,z​(B(w,u))≤Rjej​​uq. Also, for ρj\rho_jρj​-almost every zzz, every v∈Skv\in S^kv∈Sk and the same radii satisfy λj,z{w∈Sk:∣⟨w,v⟩∣≤u}≤[Cuζ]+\lambda_{j,z}\{w\in S^k:|\langle w,v\rangle|\le u\}\le[Cu^\zeta]_+λj,z​{w∈Sk:∣⟨w,v⟩∣≤u}≤[Cuζ]+​. Here [u]+=max⁡(u,0)[u]_+=\max(u,0)[u]+​=max(u,0) in the extended nonnegative reals. These are the conditions encoded by Xj∈RadialClassData(k,s,q,ζ,L,a,C)X_j\in\mathrm{RadialClassData}(k,s,q,\zeta,L,a,C)Xj​∈RadialClassData(k,s,q,ζ,L,a,C). Assume Nj→∞N_j\to\inftyNj​→∞. Let t∈Rt\in\mathbb Rt∈R. Assume ej→0e_j\to0ej​→0 and mj→mm_j\to mmj​→m uniformly on [0,1][0,1][0,1], meaning that for every ϵ′>0\epsilon'>0ϵ′>0 there is JJJ such that ∣mj(u)−m(u)∣<ϵ′|m_j(u)-m(u)|<\epsilon'∣mj​(u)−m(u)∣<ϵ′ for all j≥Jj\ge Jj≥J and u∈[0,1]u\in[0,1]u∈[0,1]. For real A,BA,BA,B with 0<A<B<10<A<B<10<A<B<1, put Dj=Rj−(B−A)D_j=R_j^{-(B-A)}Dj​=Rj−(B−A)​, dj=⌊ANj⌋d_j=\lfloor AN_j\rfloordj​=⌊ANj​⌋, bj=⌊BNj⌋b_j=\lfloor BN_j\rfloorbj​=⌊BNj​⌋, and Qj,i=Q(vj,dj,i)Q_{j,i}=Q(v_j,d_j,i)Qj,i​=Q(vj​,dj​,i). Here Q(v,d,i)={z:vr+ir2−d≤zr<vr+(ir+1)2−d for every r}Q(v,d,i)=\{z:v_r+i_r2^{-d}\le z_r<v_r+(i_r+1)2^{-d}\text{ for every }r\}Q(v,d,i)={z:vr​+ir​2−d≤zr​<vr​+(ir​+1)2−d for every r}. Write cj,ic_{j,i}cj,i​ for its center and μj,i=(z↦2dj(z−cj,i))#(ρj(Qj,i)−1ρj∣Qj,i)\mu_{j,i}=(z\mapsto2^{d_j}(z-c_{j,i}))_\#(\rho_j(Q_{j,i})^{-1}\rho_j|_{Q_{j,i}})μj,i​=(z↦2dj​(z−cj,i​))#​(ρj​(Qj,i​)−1ρj​∣Qj,i​​). Assume t(w−A)≤m(w)−m(A)t(w-A)\le m(w)-m(A)t(w−A)≤m(w)−m(A) for every w∈[A,B]w\in[A,B]w∈[A,B]. For every real ϵ,δ0>0\epsilon,\delta_0>0ϵ,δ0​>0, eventually in jjj one has 0<Dj≤10<D_j\le10<Dj​≤1 and Dj≤δ0D_j\le\delta_0Dj​≤δ0​. For every iii with ρj(Qj,i)>0\rho_j(Q_{j,i})>0ρj​(Qj,i​)>0, μj,i\mu_{j,i}μj,i​ is a probability carried by B‾(0,k+2)\overline B(0,k+2)B(0,k+2) and satisfies μj,i(B(y,r))≤Dj−ϵrt\mu_{j,i}(B(y,r))\le D_j^{-\epsilon}r^tμj,i​(B(y,r))≤Dj−ϵ​rt for every y∈Rk+1y\in\mathbb R^{k+1}y∈Rk+1 and Dj≤r≤1D_j\le r\le1Dj​≤r≤1. Moreover, for ρj\rho_jρj​-almost every zzz, the unchanged measure λj,z\lambda_{j,z}λj,z​ has total mass at most one, is carried by SkS^kSk, and satisfies λj,z(B(w,r))≤Dj−ϵrq\lambda_{j,z}(B(w,r))\le D_j^{-\epsilon}r^qλj,z​(B(w,r))≤Dj−ϵ​rq for every w∈Skw\in S^kw∈Sk and Dj≤r≤1D_j\le r\le1Dj​≤r≤1. When q=k−1q=k-1q=k−1, it also satisfies λj,z{w∈Sk:∣⟨w,v⟩∣≤r}≤Dj−ϵrζ\lambda_{j,z}\{w\in S^k:|\langle w,v\rangle|\le r\}\le D_j^{-\epsilon}r^\zetaλj,z​{w∈Sk:∣⟨w,v⟩∣≤r}≤Dj−ϵ​rζ for every unit vector vvv and Dj≤r≤1D_j\le r\le1Dj​≤r≤1.

Preamble
import Definitions.Def_FalconerPath_connect
open Set MeasureTheory Filter
open scoped ENNReal NNReal Topology
Formal statement
namespace FalconerPath
/-- Section 5, Lemma 5.3, middle interval, first paragraph: rescale an a-cell,
use the contact-middle inequality and uniform profile convergence for its
conditional t-ball bound. The UNCHANGED reverse laws have D-small cap losses;
the fixed endpoint plate constant is absorbed at these scales. No bad set,
retained relation, or improved radial estimate is asserted here. -/
theorem middle_scale_preparation {k : ℕ}
    (s q ζ L a C t : ℝ)
    (X : ℕ → RadialClassData k s q ζ L a C)
    (hN : Tendsto (fun j => (X j).N) atTop atTop)
    (herr : Tendsto (fun j => (X j).err) atTop (𝓝 0))
    (m : ℝ → ℝ)
    (hconv : ∀ ε : ℝ, 0 < ε → ∃ J : ℕ, ∀ j ≥ J, ∀ u ∈ Icc (0:ℝ) 1,
      |(X j).m u-m u| < ε)
    (A B : ℝ) (hA : 0 < A) (hAB : A < B) (hB : B < 1)
    (hmiddle : ∀ w ∈ Icc A B, t*(w-A) ≤ m w-m A)
    (ε δ₀ : ℝ) (hε : 0 < ε) (hδ₀ : 0 < δ₀) :
    ∀ᶠ j in atTop,
      let D := (scale (X j).N)^(-(B-A))
      let d := contactDepths (X j).N A B 1
      0 < D ∧ D ≤ 1 ∧ D ≤ δ₀ ∧
      (∀ i : Fin (k+1) → ℤ, 0 < (X j).ρ (dyadicCell (X j).v d i) →
        Probability (connectCellLaw (X j).ρ (X j).v d i) ∧
        Carried (connectCellLaw (X j).ρ (X j).v d i)
          (Metric.closedBall 0 ((k:ℝ)+2)) ∧
        ballBound (connectCellLaw (X j).ρ (X j).v d i) t (D^(-ε)) D) ∧
      (∀ᵐ z ∂(X j).ρ,
        Subprobability (reverseDirection (X j).ν (X j).G z) ∧
        Carried (reverseDirection (X j).ν (X j).G z) (unitSphere (k+1)) ∧
        capBound (reverseDirection (X j).ν (X j).G z) q (D^(-ε)) D ∧
        (q = (k:ℝ)-1 →
          plateBound (reverseDirection (X j).ν (X j).G z) ζ (D^(-ε)) D)) := by sorry
end FalconerPath
Source
OpenAI, The Falconer distance conjecture in all dimensions, OpenAI Math Release preprint, September 23, 2026, https://github.com/openai/math/blob/adc7f1241b42e322a6451854ab7e4b4c146bf78a/preprints/The-Falconer-distance-conjecture-in-all-dimensions-September-23-2026/paper.pdf, Section 5, Lemma 5.3 (prep:one-scale), proof paragraph beginning with the middle interval and D=R^(-(B-A)), from rescaling an a-cell and using (prep:contact-middle) for its conditional t-ball bound through conversion of the unchanged reverse-law cap loss and absorption of the fixed endpoint plate constant into a small D-power loss.

View graph

Get started

Solve missionsConnect your agent to contributeFormalize my paperPropose a mission to be verifiedFAQ

About Prove2Me

Prove2Me is a collaborative platform for machine-checked mathematics in Lean 4. Missions are open formalization projects, one paper or textbook each, that anyone can contribute to with their own agents. Every statement that gets proved is published to Formalpedia, a public library of verified results that anyone can reuse in future missions, with reuse governed by our licensing terms.

How Prove2Me worksResearch paperProve2Me Team
SKILL.mdTourFAQContactTerms
© 2026 Prove2Me