Bounded model from a structured part and mixed-progression estimates
ProvedGreenTao.bounded_model_of_mixed_estimatesadditive-combinatoricsgowers-uniformitygreen-tao
Let , let for a positive integer , and let . Suppose and
For every nonempty subset , let when and otherwise. Assume
where is the multilinear progression average. Then there exists such that
The count includes zero common differences. This finite implication isolates the normalization and multilinearity step in passing from a structured decomposition to a bounded model. It needs no pseudorandomness assumption, no primality assumption, and no asymptotic limit.
Preamble
import Definitions.Def_GreenTao_Uniformity open scoped BigOperators open GreenTao
Formal statement
theorem GreenTao.bounded_model_of_mixed_estimates (k : ℕ) (m : ℕ+) (f s u : ZMod (m : ℕ) → ℝ)
(ε ξ : ℝ) (hε : 0 ≤ ε) (hξ : 0 ≤ ξ)
(hs : ∀ x, 0 ≤ s x ∧ s x ≤ 1 + ε)
(htu : ∀ x, 0 ≤ s x + u x ∧ s x + u x ≤ f x)
(hmean : avg f - ε ≤ avg s)
(hmixed : ∀ S : Finset (Fin k), S.Nonempty →
|mixedAPAvg k (fun j x => if j ∈ S then u x else s x)| ≤ ξ) :
∃ g : ZMod (m : ℕ) → ℝ,
(∀ x, 0 ≤ g x ∧ g x ≤ 1) ∧
avg f - 2 * ε ≤ avg g ∧
apAvg k g - (2 : ℝ) ^ k * ξ ≤ apAvg k f := by sorrySource
Green and Tao, The primes contain arbitrarily long arithmetic progressions, https://arxiv.org/html/math/0404188v6, §8 proof of Theorem 3.5 assuming Proposition 8.1: the mixed-progression expansion and monotonicity for the truncated function, together with footnote 16 for normalization to [0,1]. This auxiliary finite formulation makes the elementary errors 2ε and 2^k ξ explicit; it is not a verbatim numbered theorem.