Generalized von Neumann estimate with uniform cube thresholds
ProvedGreenTao.generalized_von_neumannLet , let be prime moduli tending to infinity, and let be a -pseudorandom family. For every there is such that, for all sufficiently large , the following holds uniformly over all tuples of real functions on :
imply
Here is the real Gowers cube average. One sufficiently uniform factor makes a mixed progression average small. The threshold is chosen before the modulus and the tuple of functions; this uniformity is needed to apply the estimate to functions supplied by a structured decomposition.
Formalization Note. This is the qualitative threshold form of Proposition 5.3 for consecutive coefficients . Its additive constant is in place of , obtained by the fixed convex-mixture/rescaling allowed by Lemma 3.4. Smallness is stated in terms of the -th power of the Gowers norm.
import Definitions.Def_GreenTao_Uniformity open Filter
theorem GreenTao.generalized_von_neumann
(k : ℕ) (hk : 3 ≤ k) (M : ℕ → ℕ+)
(hprime : ∀ n, Nat.Prime (M n : ℕ))
(hM : Tendsto (fun n => (M n : ℕ)) atTop atTop)
(ν : GreenTao.Family M) (hν : GreenTao.Pseudorandom k M ν)
(η : ℝ) (hη : 0 < η) :
∃ ρ : ℝ, 0 < ρ ∧ ∀ᶠ n in atTop,
∀ F : Fin k → ZMod (M n : ℕ) → ℝ,
(∀ j x, |F j x| ≤ ν n x + 2) →
(∃ j, |GreenTao.cubeAvg (k - 1) (F j)| ≤ ρ) →
|GreenTao.mixedAPAvg k F| ≤ η := by sorry