Theorem 4.2 (= Theorem 6.4): cubic–quintic upper bound
OpenGrothendieckLB.Supplement.theorem_4_2gaussian-analysisgrothendieck-constantkrivine-roundingp2o-gran-per-chapterp2o-plan-paperp2o-v1
Let be the correlation function of the cubic–quintic limiting scheme (Definition 4.1) with parameters
and let be the inverse majorant of , where is its local inverse at the origin. At we have . Consequently
This is the paper's upper bound: it improves Krivine's bound by more than . Theorem 6.4 (p.29) is the same statement without the last inequality.
Formalization Note All decimals are exact rational constants. The local inverse of is asserted to exist, and (with convergence of the series) is asserted for every local inverse branch, all of which share the same germ. is the published grothendieckConst.
Differences from the source:
- reading: Let M be the inverse majorant of H (presupposes the local inverse of H) → existence of a local inverse branch is a conjunct; M(γ) < 1 holds for every local inverse branch; reason: makes the presupposition explicit, so the M(γ) clause cannot hold vacuously; p.26 states the inverse exists because H′(0) = b₁ > 0
Preamble
import Mathlib import Definitions.Def_GrothendieckConstantDefs import Definitions.Def_GrothendieckLB_IsLocalInverseBranch import Definitions.Def_GrothendieckLB_taylorCoeff import Definitions.Def_GrothendieckLB_Supplement_cqCorr
Formal statement
namespace GrothendieckLB
namespace Supplement
/-- Theorem 4.2 (= Theorem 6.4), cubic–quintic upper bound (Saha et al., *New Lower and Upper Bounds for the Grothendieck Constant*, arXiv:2608.11158 (held version, sha256 e7eb5d65…), Theorem 4.2, p.24; Theorem 6.4,
(32), p.29): let `H` be the correlation function of the cubic–quintic limiting scheme with the
parameters (10), `η = 0.136419125`, `s₃ = 0.34101124`, `s₅ = 0.05276111`, and `M` the inverse
majorant of `H`. At `γ = 0.881545409`, `M(γ) < 1`. Consequently
`K_G ≤ π/(2γ) ≤ 1.7818666069360661 < π/(2 log(1 + √2)) − 3.4737 · 10⁻⁴`.
**Formalization Note** `H = cqCorr η s₃ s₅` (Definition 4.1, (13)). The local inverse exists
(`H′(0) = b₁ > 0`, p.26) and is stated to exist; for every local inverse branch `G`
(base `IsLocalInverseBranch`, unique germ) with `a_n = taylorCoeff n G`, the majorant
`M(γ) = ∑_{n ≥ 1} |a_n| γ^n` (22) converges and is `< 1`. All decimals are exact rationals.
`K_G` is the published `GrothendieckConstant.grothendieckConst`. -/
theorem theorem_4_2 :
(∃ G : ℝ → ℝ, IsLocalInverseBranch (cqCorr 0.136419125 0.34101124 0.05276111) G) ∧
(∀ G : ℝ → ℝ, IsLocalInverseBranch (cqCorr 0.136419125 0.34101124 0.05276111) G →
Summable (fun n : ℕ => |taylorCoeff (n + 1) G| * (0.881545409 : ℝ) ^ (n + 1)) ∧
∑' n : ℕ, |taylorCoeff (n + 1) G| * (0.881545409 : ℝ) ^ (n + 1) < 1) ∧
GrothendieckConstant.grothendieckConst ≤ Real.pi / (2 * 0.881545409) ∧
Real.pi / (2 * 0.881545409) ≤ (1.7818666069360661 : ℝ) ∧
(1.7818666069360661 : ℝ) <
Real.pi / (2 * Real.log (1 + Real.sqrt 2)) - 3.4737e-4 := by sorry
end Supplement
end GrothendieckLB
Source
Saha et al., New Lower and Upper Bounds for the Grothendieck Constant, arXiv:2608.11158 (held version, sha256 e7eb5d65…), Theorem 4.2, p.24; Theorem 6.4, (32), p.29