Theorem 3.9: from a limiting Krivine scheme
OpenGrothendieckLB.Supplement.theorem_3_9gaussian-analysisgrothendieck-constantkrivine-roundingp2o-gran-per-chapterp2o-plan-paperp2o-v1
Let be a limiting Krivine scheme with correlation function , and suppose , so that has a local holomorphic inverse at the origin. Write
If lies within the radius of convergence of and , then
This is the upper-bound machinery of the paper: a limiting scheme with a good inverse majorant bounds the Grothendieck constant.
Formalization Note is the published grothendieckConst. The local inverse is the base IsLocalInverseBranch with Taylor coefficients ; " within the radius of convergence" is for some .
Differences from the source:
- reading: γ > 0 within the radius of convergence of H⁻¹ → ∑ |a_n| R^n converges for some R > γ (γ strictly less than the radius); reason: 'within' read as strictly inside the disk of convergence, which is what the proof (Lemma 3.10, 0 < γ < R₀) uses
Preamble
import Mathlib import Definitions.Def_GrothendieckConstantDefs import Definitions.Def_GrothendieckLB_IsLocalInverseBranch import Definitions.Def_GrothendieckLB_taylorCoeff import Definitions.Def_GrothendieckLB_Supplement_LimitingScheme_corr
Formal statement
namespace GrothendieckLB
namespace Supplement
/-- Theorem 3.9, upper bound from a limiting Krivine scheme (Saha et al., *New Lower and Upper Bounds for the Grothendieck Constant*, arXiv:2608.11158 (held version, sha256 e7eb5d65…), Theorem 3.9, p.21): let
`S = (f, g, ρ₁, …, ρ_k)` be a limiting Krivine scheme with correlation function `H`, and suppose
`H′(0) ≠ 0`, so that `H` has a local holomorphic inverse at the origin. Write
`H⁻¹(z) = ∑_{j ≥ 0} a_{2j+1} z^{2j+1}` and `M(r) = ∑_{j ≥ 0} |a_{2j+1}| r^{2j+1}`. If there exists
`γ > 0`, within the radius of convergence of `H⁻¹`, such that `M(γ) = 1`, then `K_G ≤ π/(2γ)`.
**Formalization Note** `K_G` is the published `GrothendieckConstant.grothendieckConst`. `G` is the
local inverse branch (base `IsLocalInverseBranch`) and `a_n = taylorCoeff n G`. "`γ` within the
radius of convergence" is `∑ |a_n| R^n < ∞` for some `R > γ` (i.e. `γ` strictly inside the disk of
convergence); it makes the series `M(γ)` converge. `M` runs over odd indices, as printed. -/
theorem theorem_3_9 {k : ℕ} (S : LimitingScheme k) (hderiv : deriv S.corr 0 ≠ 0) (G : ℝ → ℝ)
(hG : IsLocalInverseBranch S.corr G) (γ : ℝ) (hγ : 0 < γ)
(hrad : ∃ R : ℝ, γ < R ∧ Summable (fun n => |taylorCoeff n G| * R ^ n))
(hM : ∑' j : ℕ, |taylorCoeff (2 * j + 1) G| * γ ^ (2 * j + 1) = 1) :
GrothendieckConstant.grothendieckConst ≤ Real.pi / (2 * γ) := 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 3.9, p.21