Prove2Me
Navigate
DiscoverCollectionsFormalpediaGalaxyBlogsUsersMomentumMy Missions+
Prove2Me
Log in
← Formalpedia

Claim 6.2 — splitting g∈HIBℓg\in H^{B_\ell}_Ig∈HIBℓ​​ into ≤p(C)\le p(C)≤p(C) factors of block size ≤(b−a)/C\le(b-a)/C≤(b−a)/C with ≤p(C)\le p(C)≤p(C) relations

Proved
PolySwap.KMS.claim_6_2

by Shuze Chen · Nov 11, 2026 · Mathlib 0df444a (Lean v4.33.1)

high-dimensional-expandersp2o-gran-per-chapterp2o-plan-paperp2o-v1swap-coboundary-expansion

Let I\mathcal II be a nice set of TTT disjoint colour sets of size r+1r+1r+1 in [d][d][d], J∈IJ\in\mathcal IJ∈I, and work in JJJ's path order. There is a universal polynomial ppp (independent of ℓ\ellℓ, rrr and the bin [a,b][a,b][a,b]) such that for every 0≤ℓ≤r0\le\ell\le r0≤ℓ≤r with Bℓ=[a,b]B_\ell=[a,b]Bℓ​=[a,b], every C∈NC\in\mathbb NC∈N with 1≤C≤201\le C\le201≤C≤20, every I∈I∖{J}I\in\mathcal I\setminus\{J\}I∈I∖{J} and every g∈HIBℓg\in H^{B_\ell}_Ig∈HIBℓ​​ there are g1,…,gs∈HIBℓg_1,\dots,g_s\in H^{B_\ell}_Ig1​,…,gs​∈HIBℓ​​ with s≤p(C)s\le p(C)s≤p(C), g=g1⋯gsg=g_1\cdots g_sg=g1​⋯gs​ and each gig_igi​ supported on block size ≤b−aC\le\frac{b-a}{C}≤Cb−a​. Moreover ggg can be transformed into g1⋯gsg_1\cdots g_sg1​⋯gs​ by applying at most p(C)p(C)p(C) relations of RIℓR^\ell_IRIℓ​, the multiplication-table relations of HIBℓH^{B_\ell}_IHIBℓ​​.

Differences from the source:

  • encoding: a universal polynomial p (independent of ℓ, r, [a, b]) → ∃ p : Polynomial ℕ before d, r, T, F, I, J, ℓ; reason: universal: independent of every parameter (the proof gives p(C) = 6²C¹⁸)
  • encoding: g can be transformed into g₁⋯g_s by applying at most p(C) relations in R^ℓ_I → area (multTable (binH …) I) (cosetWord g · ([g₁]⋯[g_s])⁻¹) ≤ p(C) in the free group on the generators S_ℓ = ⋃H^{B_ℓ}_I ∖ {1} of Y_ℓ, where cosetWord g is the letter [g] (the empty word for g = 1, which is not a generator); reason: a word can be deduced from another with k relation applications (free moves not counted) iff their quotient has area ≤ k
  • encoding: a prime power q = p^m and the field F_q → a finite field F (Fintype F) with q = Fintype.card F and p = ringChar F; reason: every finite field has prime-power order q and prime characteristic p; same instances
  • encoding: a nice set I ∈ ([d] choose r+1)^[T] of disjoint colours (§6, §7 standing) and J ∈ I → I ∈ disjointColorSets d r T, IsNice d I, J ∈ I; bins indexed by ℓ ≤ r with B_ℓ = [binLo, binHi]; reason: the section's standing assumptions as binders
  • reading: the vertex link (F^I X){gH_J} ≅ CC(U{[d]∖J}; (U_{[d]∖(J∪I)}){I∈I∖{J}}) in the upper-triangular model of §3.3, with the colours cyclically relabelled so that a chosen j₀ ∈ J becomes 0 (Definition 3.6) → linkY / binY: subgroups UL d F B = ⟨e{i,i+1}(λ) : i ∈ B⟩ of SL_{d+1}(F_q) with the relabelling j₀ = min J (base J); reason: Definition 3.6 lets j₀ ∈ J be chosen and states the conditions do not depend on the choice; the least element is one such choice
  • reading: condition (e) of Definition 3.6: 'for every pair of J-bins B, B′' → IsNice quantifies (e) over every pair (ℓ, ℓ′) of nonempty bins, B = B′ included; empty bins are omitted, as the definition says; reason: a pair is an arbitrary pair, as the proof of Lemma 3.7 counts it (p. 34: 'This count allows the two bins to belong to arbitrary color sets, as in the definition'); reading it as two different bins would weaken Lemma 3.7
  • reading: Definition 2.35: generators 'all the elements of the H_i', relations R_i 'all the nontrivial relations in the multiplication table of H_i of length at most 3' (Appendix B, p. 61, writes S = ⋃H_i ∖ {1}) → cosetDehn: letters = the elements of ⋃_i H_i ∖ {1}; R_i = the positive words [a][b], [a][b][c] with letters in H_i ∖ {1} whose product is 1; reason: Appendix B (pp. 61–62), where Theorem 3.19 is proved, fixes S = ⋃H_i ∖ {1} and uses exactly the relations [g][g⁻¹] = 1 and the three-letter triangle words
  • encoding: Dehn functions, areas, radii and diameters with values in ℕ (a max over a finite set) → values in ℕ∞ (iSup/iInf; ⊤ when there is no filling or path); reason: no junk value 0 for an empty or unbounded set; equal to the paper's value whenever that is finite
Preamble
import Mathlib
import Definitions.Def_PolySwap_KMS_Complex
import Definitions.Def_PolySwap_KMS_Dehn
import Definitions.Def_PolySwap_KMS_KMSComplex
Formal statement
namespace PolySwap.KMS

/-- **Claim 6.2**, p. 47. There is a universal polynomial `p` (independent of `ℓ`, `r`, the bin
`[a, b]`, and of `d`, `q`, `T`, `I`, `J`) such that: for every nice set `I` of disjoint colour
sets, every `J ∈ I`, every `0 ≤ ℓ ≤ r` with `B_ℓ = [a, b]`, every `C ∈ ℕ` with `1 ≤ C ≤ 20`,
every `I' ∈ I \ {J}` and every `g ∈ H^{B_ℓ}_{I'}`, there are `g₁, …, g_s ∈ H^{B_ℓ}_{I'}` with
`s ≤ p(C)`, `g = g₁ ⋯ g_s`, each `g_i` supported on block size `≤ (b - a)/C`, and `g` can be
transformed into `g₁ ⋯ g_s` by at most `p(C)` relations of `R^ℓ_{I'}`. -/
theorem claim_6_2 : ∃ p : Polynomial ℕ, ∀ (d r T : ℕ) (F : Type*) [Field F] [Fintype F]
    (I : Finset (Finset ℕ)), I ∈ disjointColorSets d r T → IsNice d I →
    ∀ J ∈ I, ∀ ℓ ≤ r, ∀ C : ℕ, 1 ≤ C → C ≤ 20 → ∀ K : {K // K ∈ I.erase J},
    ∀ g, (hg : g ∈ binH d F I J ℓ K) →
      ∃ gs : List (CosetGen (binH d F I J ℓ)), gs.length ≤ p.eval C ∧
        (∀ x ∈ gs, x.1 ∈ binH d F I J ℓ K) ∧ (gs.map Subtype.val).prod = g ∧
        (∀ x ∈ gs, ∃ k : ℕ, (k : ℝ) ≤ ((binHi d J ℓ : ℝ) - binLo d J ℓ) / C ∧
          SupportedOnBlock d F
            (((x.1 : UL d F (bin d J ℓ)) : Matrix.SpecialLinearGroup (Fin (d + 1)) F) :
              Matrix (Fin (d + 1)) (Fin (d + 1)) F) k) ∧
        area (multTable (binH d F I J ℓ) K)
            (cosetWord (binH d F I J ℓ) g ⟨K, hg⟩ *
              ((gs.map FreeGroup.of).prod)⁻¹) ≤ ((p.eval C : ℕ) : ℕ∞) := by sorry

end PolySwap.KMS
Source
Dikstein, Oppenheim, Valentiner-Branth, Polynomial Swap Coboundary Expansion with Applications to LTCs and PCPs, STOC 2027 submission (manuscript sha256 b9cb2095…), Claim 6.2, p. 47
Human review
  • Endorsed by Community (Bot) · Nov 11, 2026

    Confirmed by the moderator at approval.

  • Endorsed by Shuze Chen · Nov 11, 2026

    Confirmed by the mission captain (proposal self-audit).

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